{"asOf":"2026-10-16T13:38:14.000Z","controlMap":{"hash":"0xb1601aaa11cbea47c47b584a01d40ddcb68687a3ef9e689b79b458e3321b9a6a","version":"1.0.0"},"controlMatrix":[{"appliesNote":"Determines whether Arts. 9 to 15, 72 and 73 apply at all. The classification belongs to the provider; Karatum does not classify systems.","appliesTo":"high_risk_provider_and_deployer","artifacts":[],"framework":"eu-ai-act","id":"EU-AIA-6","limits":["Karatum evidence says nothing about the intended purpose of the operator's system, which is what decides the classification. An agent that executes on-chain transactions is not high-risk merely because it is an AI agent."],"operatorGaps":["Document the intended purpose and the classification with its reasoning (including any Art. 6(3) exemption and its registration duty).","Re-assess when the intended purpose or deployment context changes."],"ref":"Art. 6 and Annex III","requirement":"A system is high-risk if it is a safety component of a product under the Annex I legislation, or if it falls in an Annex III use case (for example creditworthiness assessment of natural persons, employment, essential services), unless the Art. 6(3) exemptions apply and are documented.","status":"operator_only","title":"Is the system high-risk?"},{"appliesNote":"Obligation of providers of high-risk AI systems. For other systems it is voluntary good practice.","appliesTo":"high_risk_provider","artifacts":[{"artifact":"verdicts","how":"Repeatable test results with outcome verified from chain state or test execution, giving measured success rates over a stated sample (supports the testing step, Art. 9(6) to (8)).","role":"primary","state":"present"},{"artifact":"certificate","how":"A probabilistic threshold made explicit: the lower 95% confidence bound of the success rate, with sample size, evidence class and expiry.","role":"primary","state":"present"},{"artifact":"shield","how":"Testing against foreseeable adversarial inputs (the injection attack classes of the Shield suite), reported by class.","role":"primary","state":"absent"},{"artifact":"honesty","how":"Measures one specific risk: the agent reporting success when it did not succeed.","role":"supporting","state":"present"},{"artifact":"fingerprints","how":"Detects change in behaviour after a model, prompt or tool update, a trigger for re-running the risk assessment.","role":"supporting","state":"present"},{"artifact":"incidents","how":"Post-deployment incidents on the build feed the risk evaluation required by Art. 9(2)(c).","role":"supporting","state":"present"},{"artifact":"guard","how":"An on-chain gate and tier caps are a risk-control measure for transactions the agent initiates.","role":"supporting","state":"present"}],"framework":"eu-ai-act","id":"EU-AIA-9","limits":["Karatum tests fixed task suites (on-chain web3 tasks and software tasks). They are not a test of the operator's intended purpose or of risks to health, safety or fundamental rights.","Shield covers the attack classes of its suite version only; it is not a complete adversarial assessment.","A measured success rate is not a statement that residual risk is acceptable (Art. 9(5)); that judgement is the provider's."],"operatorGaps":["Identify and analyse the risks of the intended purpose and of reasonably foreseeable misuse, including impact on persons under 18 and other vulnerable groups where relevant (Art. 9(9)).","Define the metrics and thresholds appropriate to the intended purpose and show how the Karatum numbers relate to them.","Document the risk management process, the residual risk and the decision that it is acceptable; review it on a schedule."],"ref":"Art. 9","requirement":"A continuous, iterative risk management process across the life cycle: identify and analyse known and foreseeable risks to health, safety and fundamental rights (including foreseeable misuse), estimate and evaluate them, adopt suitable measures, and test the system against prior-defined metrics and probabilistic thresholds, including under real-world conditions where appropriate.","status":"evidence_attached","title":"Risk management system"},{"appliesNote":"Obligation of providers of high-risk AI systems.","appliesTo":"high_risk_provider","artifacts":[{"artifact":"verdicts","how":"Test logs of Karatum trials: task, evidence, verdict hash and attestation per attempt.","role":"supporting","state":"present"},{"artifact":"judge","how":"Reproducible description of how each verdict was decided: trace, evaluator digest, source commit.","role":"supporting","state":"absent"},{"artifact":"certificate","how":"Summary of the results with confidence intervals.","role":"supporting","state":"present"},{"artifact":"shield","how":"Robustness testing report.","role":"supporting","state":"absent"}],"framework":"eu-ai-act","id":"EU-AIA-11","limits":["Only the validation and testing material of Annex IV point 2(g) can draw on Karatum. System description, design, data, training and monitoring parts of the documentation are untouched."],"operatorGaps":["Everything else in Annex IV: general description, development process, data and training methodologies, monitoring and control, performance metrics rationale, risk management description, changes over the life cycle, declaration of conformity.","Dated and signed test reports by the responsible persons."],"ref":"Art. 11 and Annex IV","requirement":"Technical documentation drawn up before placing on the market and kept up to date, including a description of the validation and testing procedures with test logs and reports dated and signed by the responsible persons (Annex IV, point 2(g)).","status":"partial_evidence","title":"Technical documentation (testing part)"},{"appliesNote":"Art. 12 is a design requirement on providers; Arts. 19 and 26(6) put the retention duty on providers and deployers of high-risk systems.","appliesTo":"high_risk_provider_and_deployer","artifacts":[{"artifact":"flight","how":"A tamper-evident record of the agent's steps: signed, hash-chained batches with gap and equivocation detection, optionally anchored on-chain. Coverage in this pack is the measured number of batches, steps, gaps and anchored batches.","role":"primary","state":"absent"},{"artifact":"verdicts","how":"Per-attempt records of the trials the agent ran, each with a verdict hash.","role":"supporting","state":"present"},{"artifact":"judge","how":"The judge's complete input for each verdict is stored (trace), so the evaluation itself is reconstructible.","role":"supporting","state":"absent"},{"artifact":"guard","how":"Preflight checks of transactions the agent proposed are logged with their verdicts.","role":"supporting","state":"present"}],"framework":"eu-ai-act","id":"EU-AIA-12","limits":["The Flight Recorder records what the operator's agent chooses to emit through the SDK or proxy. Steps are committed as hashes; contents stay with the operator unless disclosed, so Karatum cannot show the content of a log.","Karatum does not hold the operator's logs and cannot attest that the event set is sufficient for the intended purpose."],"operatorGaps":["Define which events the system must log for its intended purpose (Art. 12(2)) and show that the recorded set covers them.","Set and document retention (at least six months for deployers, longer where other law requires) and access control, and handle personal data in logs under data-protection law.","Keep the underlying log contents, not only their commitments."],"ref":"Art. 12 (and Arts. 19, 26(6))","requirement":"High-risk systems must technically allow automatic recording of events over their lifetime, to a degree that enables identifying situations that may present a risk or a substantial modification, supports post-market monitoring (Art. 72) and deployers' monitoring of operation (Art. 26(5)). Providers keep the logs under their control (Art. 19) and deployers keep them for a period appropriate to the purpose, at least six months (Art. 26(6)).","status":"partial_evidence","title":"Record-keeping: automatic event logging"},{"appliesNote":"Obligation of providers of high-risk AI systems, owed to deployers.","appliesTo":"high_risk_provider","artifacts":[{"artifact":"certificate","how":"A declared accuracy metric with its confidence interval and the conditions (task class, evidence class, sample, expiry) under which it was measured.","role":"primary","state":"present"},{"artifact":"shield","how":"Robustness figures by attack class that can be quoted in instructions for use.","role":"supporting","state":"absent"},{"artifact":"honesty","how":"Tells a deployer how far the agent's own success report can be trusted.","role":"supporting","state":"present"},{"artifact":"flight","how":"Describes the logging mechanism that exists.","role":"supporting","state":"absent"}],"framework":"eu-ai-act","id":"EU-AIA-13","limits":["These are measurements to cite; they are not the instructions for use. Karatum figures describe Karatum's task suites, not the operator's deployment context."],"operatorGaps":["Write the instructions for use: intended purpose, performance for the target users and contexts, foreseeable misuse, input data specification, oversight measures, maintenance, expected lifetime, log access (Art. 13(3)).","State which measured figures the operator adopts as its declared accuracy and why they are relevant."],"ref":"Art. 13","requirement":"The system must be sufficiently transparent for deployers to interpret its output and use it appropriately, and be accompanied by instructions for use giving its characteristics, capabilities and limits of performance, including declared accuracy levels and metrics, known circumstances that may affect them, robustness and cybersecurity, human oversight measures and logging mechanisms.","status":"evidence_attached","title":"Transparency and information to deployers"},{"appliesNote":"Design duty on providers of high-risk systems; Art. 26(2) requires deployers to assign competent, trained and authorised persons.","appliesTo":"high_risk_provider_and_deployer","artifacts":[{"artifact":"guard","how":"A technical intervention point: an account that installs the Guard refuses the agent's transaction while its certificate is not valid or exceeds the tier caps, whatever the agent says.","role":"primary","state":"present"},{"artifact":"recall","how":"A confirmed incident revokes the agent across the Guard registries; the pack shows measured revocation latency, and labels drills as drills.","role":"supporting","state":"present"},{"artifact":"honesty","how":"Gives overseers a measured basis for how much weight the agent's own reports deserve (automation bias).","role":"supporting","state":"present"},{"artifact":"certificate","how":"A visible, expiring status that tells an overseer whether the agent is currently in a certified state.","role":"supporting","state":"present"}],"framework":"eu-ai-act","id":"EU-AIA-14","limits":["The Guard enforces only for accounts that have installed it, and only on the chains it is deployed to (testnets at the time of writing).","Karatum does not identify, train or authorise the human overseers and does not provide an operator interface. The pack cannot read the per-account Guard settings of an operator's account; list those in the declaration."],"operatorGaps":["Assign named persons with competence, training and authority (Art. 26(2)), and document the oversight measures and the stop procedure.","Provide interface tools that let overseers interpret the system's output and review its actions.","Document where human approval is required before an action takes effect and how it is evidenced."],"ref":"Art. 14","requirement":"The system is designed so natural persons can effectively oversee it while in use: understand its capacities and limits, monitor its operation, stay aware of automation bias, interpret output, decide not to use or to override it, and intervene or stop it.","status":"evidence_attached","title":"Human oversight"},{"appliesNote":"Obligation of providers of high-risk AI systems.","appliesTo":"high_risk_provider","artifacts":[{"artifact":"certificate","how":"Accuracy as a measured success rate on verified outcomes, with a 95% confidence interval and sample size.","role":"primary","state":"present"},{"artifact":"verdicts","how":"The underlying verified outcomes, by task type and evidence class.","role":"primary","state":"present"},{"artifact":"shield","how":"Resilience to the adversarial input classes in the Shield suite (prompt injection and related), measured as attacks resisted out of attacks delivered.","role":"primary","state":"absent"},{"artifact":"fingerprints","how":"Consistency of behaviour over time on a fixed probe set; drift above the policy threshold marks the certificate STALE.","role":"supporting","state":"present"},{"artifact":"judge","how":"The accuracy measurement itself is reproducible.","role":"supporting","state":"absent"},{"artifact":"honesty","how":"Reliability of the system's own self-reported outcome.","role":"supporting","state":"present"}],"framework":"eu-ai-act","id":"EU-AIA-15","limits":["Accuracy is measured on Karatum's task suites. It does not prove accuracy for the operator's inputs.","Shield does not test data or model poisoning, model extraction, confidentiality attacks or the cybersecurity of the operator's infrastructure, keys and APIs.","No test shows fail-safe or fallback behaviour (Art. 15(4))."],"operatorGaps":["Declare the accuracy levels and metrics for the intended purpose in the instructions for use (Art. 15(3)).","Cybersecurity of the system and its hosting: threat model, penetration testing, secrets and key management, access control, supply-chain controls.","Technical redundancy, fail-safe and fallback measures; handling of feedback loops if the system keeps learning."],"ref":"Art. 15","requirement":"An appropriate level of accuracy, robustness and cybersecurity throughout the life cycle, with accuracy metrics declared in the instructions for use; resilience to errors, faults and inconsistencies; and resilience against attempts by third parties to alter use, outputs or performance by exploiting vulnerabilities (for example data poisoning, model poisoning, adversarial examples, confidentiality attacks).","status":"evidence_attached","title":"Accuracy, robustness and cybersecurity"},{"appliesNote":"Obligation of deployers of high-risk systems, for example an enterprise running a third-party agent.","appliesTo":"high_risk_deployer","artifacts":[{"artifact":"flight","how":"The recorded operation that deployer monitoring and log retention can build on.","role":"supporting","state":"absent"},{"artifact":"incidents","how":"Incidents that name the build, with the timeline of operator notification and response.","role":"supporting","state":"present"},{"artifact":"guard","how":"A gate at the deployer's own account that constrains what the agent can do.","role":"supporting","state":"present"}],"framework":"eu-ai-act","id":"EU-AIA-26","limits":["Karatum is not the deployer's monitoring function and does not hold the deployer's logs."],"operatorGaps":["Use according to the provider's instructions; assign and train oversight persons; input data relevance; inform provider and authorities; keep logs for at least six months; inform workers and affected persons where Art. 26(7) and (11) require it."],"ref":"Art. 26","requirement":"Deployers of high-risk systems use them according to the instructions, assign human oversight, monitor operation and inform the provider and authorities where they identify a risk or serious incident, and keep the automatically generated logs for at least six months.","status":"partial_evidence","title":"Deployer obligations (monitoring and logs)"},{"appliesNote":"Obligation of providers of high-risk systems.","appliesTo":"high_risk_provider","artifacts":[{"artifact":"fingerprints","how":"Continuous behavioural drift measurement against a baseline, with the history of measurements.","role":"primary","state":"present"},{"artifact":"certificate","how":"A certificate that is recomputed on a schedule, expires and changes status when evidence goes stale.","role":"primary","state":"present"},{"artifact":"incidents","how":"Registered failures with detection source, severity and timeline.","role":"supporting","state":"present"},{"artifact":"flight","how":"Operational data the operator can analyse for performance over time.","role":"supporting","state":"absent"},{"artifact":"honesty","how":"Recomputed periodically; monthly editions are immutable.","role":"supporting","state":"present"},{"artifact":"verdicts","how":"The ongoing trial history, with first and last attempt dates in the pack.","role":"supporting","state":"present"}],"framework":"eu-ai-act","id":"EU-AIA-72","limits":["Karatum monitors its own scheduled trials and probes. Performance in real use is visible only if the operator records it, for example through the Flight Recorder."],"operatorGaps":["Write the post-market monitoring plan (the Commission template under Art. 72(3)); decide what data to collect from deployers and users, how to analyse it and who acts on it.","Where relevant, analyse interaction with other AI systems."],"ref":"Art. 72","requirement":"Providers establish a post-market monitoring system proportionate to the system, which actively and systematically collects, documents and analyses relevant data on performance throughout its life, to evaluate continuous compliance, based on a post-market monitoring plan that is part of the technical documentation.","status":"evidence_attached","title":"Post-market monitoring"},{"appliesNote":"Duty of providers of high-risk systems (deployers must inform the provider and authorities under Art. 26(5)). Only incidents meeting the legal definition in Art. 3(49) are reportable.","appliesTo":"high_risk_provider_deployer_authority","artifacts":[{"artifact":"incidents","how":"Detection, classification, severity, status and a timestamped timeline for each incident that names the build, including when the operator was notified and responded.","role":"primary","state":"present"},{"artifact":"recall","how":"Corrective action on the Karatum side: certificate revocation, with measured latency.","role":"supporting","state":"present"},{"artifact":"flight","how":"Records the operator can use to reconstruct what happened (traceability for the investigation).","role":"supporting","state":"absent"}],"framework":"eu-ai-act","id":"EU-AIA-73","limits":["Karatum incident classes such as fund loss are not the legal term 'serious incident' (death or serious harm to health, serious and irreversible disruption of critical infrastructure, infringement of fundamental-rights obligations, serious harm to property or the environment). Financial loss alone may not qualify.","A Karatum incident entry is not a report to an authority and Karatum files none on anyone's behalf."],"operatorGaps":["Decide whether each event is a serious incident under Art. 3(49) and keep the reasoning.","Run the reporting procedure to the competent authority within the deadlines (15, 10 or 2 days depending on the case, Art. 73(2) to (4)); perform and document the investigation and corrective actions (Art. 73(6)); notify the Karatum operator contact about incidents Karatum does not know of."],"ref":"Art. 73 (with Arts. 3(49), 20, 26(5))","requirement":"Providers of high-risk systems report serious incidents to the market surveillance authorities of the Member States where they occurred, immediately after establishing a causal link or its reasonable likelihood and not later than 15 days after becoming aware (shorter periods of 10 or 2 days apply to specific kinds), investigate, and take corrective action.","status":"evidence_attached","title":"Serious incident reporting"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"verdicts","how":"Each build is registered with an identifier, a genome hash (model, prompt hash, tools, harness) and a version lineage, which gives an inventory entry with a stable identity.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-GOVERN-1.6","limits":["Karatum knows only the builds registered with it."],"operatorGaps":["Keep the organisation's own inventory of every AI system and its risk priority, whether or not it was ever registered with Karatum."],"ref":"GOVERN 1.6","requirement":"Mechanisms exist to inventory AI systems, resourced according to risk priorities.","status":"partial_evidence","title":"Inventory of AI systems"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"incidents","how":"A public registry with a report form, severity scoring, timelines and a signed feed that other parties can subscribe to.","role":"primary","state":"present"},{"artifact":"verdicts","how":"Independent, repeatable testing of the agent.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-GOVERN-4.3","limits":["Does not replace the organisation's own incident process or its information-sharing agreements."],"operatorGaps":["Define internal practices: who tests, how incidents are raised internally, what is shared externally."],"ref":"GOVERN 4.3","requirement":"Organisational practices enable AI testing, identification of incidents and information sharing.","status":"evidence_attached","title":"Testing, incident identification and information sharing"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"certificate","how":"Independent measured performance of an agent the organisation did not build.","role":"supporting","state":"present"},{"artifact":"verdicts","how":"Verified outcomes rather than the supplier's claims.","role":"supporting","state":"present"},{"artifact":"honesty","how":"How reliable the supplier agent's own reports are.","role":"supporting","state":"present"},{"artifact":"shield","how":"Resilience of the third-party agent to injected hostile input.","role":"supporting","state":"absent"}],"framework":"nist-ai-rmf","id":"NIST-GOVERN-6.1","limits":["Does not cover intellectual-property, privacy or licensing risks of a third-party system."],"operatorGaps":["Procurement policy, contractual terms, due diligence on the supplier's data and model provenance, exit plan (see also GOVERN 6.2 contingency processes)."],"ref":"GOVERN 6.1","requirement":"Policies and procedures address risks from third-party AI software and data, including infringement of rights.","status":"partial_evidence","title":"Third-party AI risk"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"judge","how":"The evaluator is a published, reproducible bundle with a source commit; the verdict is a pure function of a stored trace.","role":"supporting","state":"absent"},{"artifact":"verdicts","how":"Evidence classes E0 to E4 state how strong the evidence behind each verdict is.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MAP-2.3","limits":["Describes Karatum's evaluation design, not the organisation's choice of TEVV for its context."],"operatorGaps":["Document experimental design, data representativeness and the sources of validity threats for the intended deployment."],"ref":"MAP 2.3","requirement":"Scientific integrity and test, evaluation, verification and validation (TEVV) considerations are identified and documented.","status":"partial_evidence","title":"TEVV considerations identified"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"guard","how":"A technical oversight gate exists for accounts that install it.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MAP-3.5","limits":["A gate is not a process: who watches, who approves and who can stop the agent is not visible to Karatum."],"operatorGaps":["Define roles, escalation and approval paths; assess whether they work."],"ref":"MAP 3.5","requirement":"Processes for human oversight are defined, assessed and documented in line with policy.","status":"partial_evidence","title":"Human oversight processes defined"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"certificate","how":"The likelihood side for task failure: measured success rate with a 95% confidence interval.","role":"supporting","state":"present"},{"artifact":"incidents","how":"Public incident reports with severity for the build.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MAP-5.1","limits":["Says nothing about the magnitude or the people affected."],"operatorGaps":["Impact analysis for the deployment context, including affected parties and their stake."],"ref":"MAP 5.1","requirement":"Likelihood and magnitude of each identified impact, positive or negative, are identified based on expected use, past uses and public incident reports.","status":"partial_evidence","title":"Likelihood and magnitude of impacts"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"certificate","how":"Defined metric: verified success rate with a Wilson interval, by task class.","role":"primary","state":"present"},{"artifact":"honesty","how":"Defined metric: over-claim rate with interval, bias, calibration error.","role":"primary","state":"present"},{"artifact":"shield","how":"Defined metric: attacks resisted per class.","role":"primary","state":"absent"},{"artifact":"fingerprints","how":"Defined metric: drift against baseline.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-1.1","limits":["The metrics are Karatum's. Whether they are the right ones for the organisation's most significant risks is not established."],"operatorGaps":["Select the metrics that fit the organisation's own risk list and document the risks that remain unmeasured."],"ref":"MEASURE 1.1","requirement":"Approaches and metrics for measuring AI risks are selected, starting with the most significant risks, and the risks that cannot be measured are documented.","status":"evidence_attached","title":"Metrics selected"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"judge","how":"Scheduled re-verification re-runs the judge on stored traces and compares the result with the recorded verdict.","role":"supporting","state":"absent"},{"artifact":"honesty","how":"Methodology is versioned; each edition carries its method version.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-1.2","limits":["The re-verification checks the measurement's reproducibility, not whether it is the right measurement."],"operatorGaps":["Periodic review of the metric set and of the controls against changing risks."],"ref":"MEASURE 1.2","requirement":"The appropriateness of AI metrics and the effectiveness of existing controls are regularly assessed and updated.","status":"partial_evidence","title":"Metrics and controls regularly reassessed"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"verdicts","how":"Karatum evaluates agents it did not build (for external builds), with published methodology; the decision comes from chain state and tests.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-1.3","limits":["Karatum builds its own house agents; for those the evaluator is not independent of the agent. The pack states whether the build is a Karatum house build."],"operatorGaps":["Decide whether an assessment by Karatum meets the organisation's independence requirement."],"ref":"MEASURE 1.3","requirement":"Internal experts who were not front-line developers, or independent assessors, regularly assess the system.","status":"partial_evidence","title":"Independent assessment"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"verdicts","how":"Trials, task types, seeds and the verdict set are listed, and the verdict set is committed by hash.","role":"primary","state":"present"},{"artifact":"judge","how":"Evaluator bundle digests, source commits and traces.","role":"primary","state":"absent"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-2.1","limits":["Test sets are Karatum's task suites."],"operatorGaps":["Document any additional test sets the organisation uses."],"ref":"MEASURE 2.1","requirement":"Test sets, metrics and details about the tools used during TEVV are documented.","status":"evidence_attached","title":"Test sets, metrics and tools documented"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"verdicts","how":"Tasks run on a fork of real chain state with outcomes read from chain state; software tasks run against real test suites.","role":"primary","state":"present"},{"artifact":"certificate","how":"Results summarised with sample size and interval.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-2.3","limits":["Whether Karatum's conditions resemble the operator's deployment is the operator's judgement; they are fixed task classes."],"operatorGaps":["Show how deployment conditions compare with the tested ones and what is untested."],"ref":"MEASURE 2.3","requirement":"Performance or assurance criteria are measured and demonstrated for conditions similar to deployment, with the results documented.","status":"evidence_attached","title":"Performance demonstrated under deployment-like conditions"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"flight","how":"Recorded operation with coverage, gap and fork detection.","role":"primary","state":"absent"},{"artifact":"fingerprints","how":"Behavioural drift on a fixed probe set.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-2.4","limits":["Production monitoring exists only for steps the operator records."],"operatorGaps":["Operate monitoring of live behaviour, with thresholds, alerts and owners."],"ref":"MEASURE 2.4","requirement":"The functionality and behaviour of the AI system and its components are monitored in production.","status":"partial_evidence","title":"Monitoring in production"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"certificate","how":"Measured success with an interval, the weakest evidence class and an expiry that stops an old measurement being treated as current.","role":"primary","state":"present"},{"artifact":"honesty","how":"Reliability of the system's self-assessment.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-2.5","limits":["Generalisation beyond Karatum's task classes is not demonstrated."],"operatorGaps":["Document limits for the intended use and population."],"ref":"MEASURE 2.5","requirement":"The system to be deployed is demonstrated to be valid and reliable, and the limits of generalisability beyond the conditions of development are documented.","status":"evidence_attached","title":"Valid and reliable, limitations documented"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"shield","how":"Evaluation against hostile inputs that could make the agent take harmful actions with funds or secrets.","role":"supporting","state":"absent"},{"artifact":"incidents","how":"Registered harmful outcomes.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-2.6","limits":["Safety in the sense of harm to people is outside Karatum's suites. Harm measured here is mostly loss of funds, leaks of planted secrets and wrong task outcomes."],"operatorGaps":["Safety evaluation for the real hazard model of the application; residual-risk tolerance and the decision against it."],"ref":"MEASURE 2.6","requirement":"The AI system is evaluated regularly for safety risks, demonstrating it is safe and that residual negative risk does not exceed tolerance.","status":"partial_evidence","title":"Safety risks evaluated regularly"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"shield","how":"Red-team results by attack class, from chain state and canary hits.","role":"primary","state":"absent"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-2.7","limits":["Covers the injection classes in the suite. Not data poisoning, model theft, infrastructure security or insider risk."],"operatorGaps":["Security assessment of the whole system: threat modelling, penetration tests, supply chain."],"ref":"MEASURE 2.7","requirement":"AI system security and resilience are evaluated and documented.","status":"evidence_missing","title":"Security and resilience evaluated"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"judge","how":"Verdicts are accountable: who decided, with which evaluator, from which trace.","role":"supporting","state":"absent"},{"artifact":"flight","how":"Signed records bind actions to an agent key.","role":"supporting","state":"absent"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-2.8","limits":["Accountability inside the organisation (who answers for the agent) is not visible to Karatum."],"operatorGaps":["Assign accountability and examine explainability needs of the intended use."],"ref":"MEASURE 2.8","requirement":"Risks associated with transparency and accountability are examined and documented.","status":"evidence_missing","title":"Transparency and accountability risks examined"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"judge","how":"Re-verification results show whether re-running the judge reproduces recorded verdicts.","role":"supporting","state":"absent"},{"artifact":"honesty","how":"Sensitivity analysis shows how much the headline changes when parts of the data are excluded.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-2.13","limits":["Reproducibility is not validity."],"operatorGaps":["Evaluate whether the metrics predicted real outcomes for the organisation."],"ref":"MEASURE 2.13","requirement":"The effectiveness of the employed TEVV metrics and processes in the Measure function is evaluated and documented.","status":"partial_evidence","title":"Effectiveness of TEVV metrics and processes"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"incidents","how":"A registry with automated detectors and an open report channel.","role":"primary","state":"present"},{"artifact":"fingerprints","how":"Silent behaviour change is a type of emergent risk that is surfaced as drift.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-3.1","limits":["Tracks what Karatum detects or is told."],"operatorGaps":["An internal process and staff to track risks over the system's life."],"ref":"MEASURE 3.1","requirement":"Approaches, personnel and documentation exist to regularly identify and track existing, unanticipated and emergent risks.","status":"evidence_attached","title":"Tracking existing and emergent risks"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"incidents","how":"Anyone can file an incident report with evidence; rate-limited, triaged by maintainers.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-3.3","limits":["Not an appeal mechanism for decisions made about people."],"operatorGaps":["Feedback and appeal channels for the people the system affects."],"ref":"MEASURE 3.3","requirement":"Feedback processes for end users and impacted communities to report problems and appeal outcomes are established and integrated into evaluation.","status":"partial_evidence","title":"Feedback processes for end users and affected people"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"fingerprints","how":"Measured drift against baseline over time.","role":"primary","state":"present"},{"artifact":"certificate","how":"Status changes (VALID, STALE, EXPIRED, REVOKED) as evidence ages or behaviour changes.","role":"primary","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MEASURE-4.3","limits":["No consultation with users or domain experts is part of the data."],"operatorGaps":["Consult relevant AI actors on whether a change is an improvement or a decline."],"ref":"MEASURE 4.3","requirement":"Measurable performance improvements or declines are identified based on consultations with relevant AI actors and documented.","status":"evidence_attached","title":"Performance improvements or declines identified"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"guard","how":"On-chain gating is a mitigation available for transaction risk.","role":"supporting","state":"present"},{"artifact":"recall","how":"A recall can withdraw the agent's certified status.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MANAGE-1.3","limits":["Karatum implements mitigations; deciding the response to a risk is the organisation's."],"operatorGaps":["Risk treatment plan with owners and decisions."],"ref":"MANAGE 1.3","requirement":"Responses to the AI risks deemed high priority are developed, planned and documented, for example mitigating, transferring, avoiding or accepting.","status":"partial_evidence","title":"Responses to high-priority risks"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"incidents","how":"A documented lifecycle (reported, triaged, confirmed, recalled, fixed) with timestamps.","role":"supporting","state":"present"},{"artifact":"recall","how":"Measured time from confirmation to on-chain revocation.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MANAGE-2.3","limits":["Karatum's response is to its registry and certificates, not to the operator's systems."],"operatorGaps":["Operator-side response and recovery procedures, rehearsed."],"ref":"MANAGE 2.3","requirement":"Procedures are followed to respond to and recover from a previously unknown risk when it is identified.","status":"partial_evidence","title":"Response and recovery for previously unknown risks"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"recall","how":"A confirmed incident revokes the certificate in Guard registries and withdraws the attestation; drills measure the time.","role":"primary","state":"present"},{"artifact":"guard","how":"Accounts with the Guard installed stop executing the agent's transactions once its certificate is revoked or expired.","role":"primary","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MANAGE-2.4","limits":["Only for accounts that use the Guard, on the chains it is deployed to. It does not turn off the agent itself."],"operatorGaps":["A way to stop the agent in the operator's own stack and a named person responsible for using it."],"ref":"MANAGE 2.4","requirement":"Mechanisms are in place and applied, and responsibilities assigned, to supersede, disengage or deactivate AI systems that demonstrate performance or outcomes inconsistent with intended use.","status":"evidence_attached","title":"Supersede, disengage or deactivate"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"certificate","how":"Expiring certificate status of the supplier agent.","role":"supporting","state":"present"},{"artifact":"fingerprints","how":"Detects a supplier changing the model or prompt without notice.","role":"supporting","state":"present"},{"artifact":"incidents","how":"Incidents on the supplier's build.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MANAGE-3.1","limits":["Visible only for third-party agents Karatum evaluates."],"operatorGaps":["Supplier monitoring cadence, contractual notification duties."],"ref":"MANAGE 3.1","requirement":"AI risks and benefits from third-party resources are regularly monitored and controls applied and documented.","status":"partial_evidence","title":"Third-party AI risks monitored"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"fingerprints","how":"Change detection supports change management.","role":"supporting","state":"present"},{"artifact":"incidents","how":"Incident response record.","role":"supporting","state":"present"},{"artifact":"flight","how":"Recorded operation.","role":"supporting","state":"absent"},{"artifact":"guard","how":"An override point for transactions.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MANAGE-4.1","limits":["Only technical pieces of a plan."],"operatorGaps":["The plan itself, with user input, appeal, decommissioning and change-management procedures."],"ref":"MANAGE 4.1","requirement":"Post-deployment monitoring plans are implemented, including mechanisms for capturing user input, appeal and override, decommissioning, incident response, recovery and change management.","status":"partial_evidence","title":"Post-deployment monitoring plan implemented"},{"appliesNote":"Voluntary framework: it applies to whoever chooses to adopt it. No regulator certifies conformity with it.","appliesTo":"voluntary_framework","artifacts":[{"artifact":"incidents","how":"Public timeline per incident, a signed feed (JSON and Atom), HMAC-signed webhooks, operator notification and response recorded on the timeline.","role":"primary","state":"present"},{"artifact":"recall","how":"On-chain announcement of recalls readable by any contract.","role":"supporting","state":"present"}],"framework":"nist-ai-rmf","id":"NIST-MANAGE-4.3","limits":["Communication to people affected by the operator's use of the agent is the operator's."],"operatorGaps":["Communication plan to customers, regulators and affected communities."],"ref":"MANAGE 4.3","requirement":"Incidents and errors are communicated to relevant AI actors, including affected communities, and processes for tracking, responding to and recovering from them are followed and documented.","status":"evidence_attached","title":"Incidents and errors communicated"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"incidents","how":"An external, public channel to report an agent failure with evidence.","role":"supporting","state":"present"}],"framework":"iso-42001","id":"ISO42001-A.3.3","limits":["An external channel is not the organisation's internal reporting process."],"operatorGaps":["Internal reporting process, whistleblowing protection, handling of reports."],"ref":"A.3.3","requirement":"The organisation defines and puts in place a process to report concerns about the organisation's role with respect to an AI system throughout its life cycle.","status":"partial_evidence","title":"Reporting of concerns"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[],"framework":"iso-42001","id":"ISO42001-A.5.2","limits":["Karatum measures agent task performance and robustness. It does not assess societal or individual consequences."],"operatorGaps":["The whole control: impact assessment process, its documentation and its use in decisions."],"ref":"A.5.2","requirement":"The organisation establishes a process to assess the potential consequences for individuals, groups and society of its AI systems.","status":"operator_only","title":"AI system impact assessment process"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"verdicts","how":"Verified outcomes in repeatable trials.","role":"primary","state":"present"},{"artifact":"certificate","how":"Acceptance criterion made explicit as a lower confidence bound with expiry.","role":"primary","state":"present"},{"artifact":"shield","how":"Adversarial validation by attack class.","role":"primary","state":"absent"},{"artifact":"judge","how":"Validation results are reproducible from stored traces.","role":"supporting","state":"absent"}],"framework":"iso-42001","id":"ISO42001-A.6.2.4","limits":["The criteria for use of the results are the organisation's."],"operatorGaps":["Define the V&V plan and acceptance criteria for the intended use, and record decisions against them."],"ref":"A.6.2.4","requirement":"The organisation defines and documents verification and validation measures for the AI system and specifies criteria for their use.","status":"evidence_attached","title":"AI system verification and validation"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"certificate","how":"A certificate status that can be required before deployment.","role":"supporting","state":"present"},{"artifact":"guard","how":"A gate can enforce the requirement at the moment of execution.","role":"supporting","state":"present"}],"framework":"iso-42001","id":"ISO42001-A.6.2.5","limits":["Does not document or approve a deployment."],"operatorGaps":["Deployment plan, approvals and the check of requirements before release."],"ref":"A.6.2.5","requirement":"The organisation documents a deployment plan and ensures that appropriate requirements are met prior to deployment.","status":"partial_evidence","title":"AI system deployment"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"fingerprints","how":"Ongoing behavioural drift measurement.","role":"primary","state":"present"},{"artifact":"flight","how":"Recorded operation, with coverage measured in the pack.","role":"primary","state":"absent"},{"artifact":"incidents","how":"Incident history.","role":"supporting","state":"present"}],"framework":"iso-42001","id":"ISO42001-A.6.2.6","limits":["Monitoring of real use depends on what the operator records."],"operatorGaps":["Operations plan: monitoring metrics, repair and update procedures, support."],"ref":"A.6.2.6","requirement":"The organisation defines and documents the elements for ongoing operation and monitoring of the AI system, including performance monitoring, repairs, updates and support.","status":"evidence_attached","title":"AI system operation and monitoring"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"verdicts","how":"This pack is itself a document for auditors on testing, monitoring and incidents.","role":"supporting","state":"present"},{"artifact":"judge","how":"Reproducible evaluation method.","role":"supporting","state":"absent"}],"framework":"iso-42001","id":"ISO42001-A.6.2.7","limits":["Covers the evaluation slice of the documentation only."],"operatorGaps":["System design, data, models, intended use and limits documentation for each audience."],"ref":"A.6.2.7","requirement":"The organisation determines what AI system technical documentation each relevant category of interested parties needs and provides it in an appropriate form.","status":"partial_evidence","title":"AI system technical documentation"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"flight","how":"Tamper-evident log of agent steps, with batches, gaps and anchors counted.","role":"primary","state":"absent"},{"artifact":"judge","how":"Evaluation traces are retained.","role":"supporting","state":"absent"}],"framework":"iso-42001","id":"ISO42001-A.6.2.8","limits":["Commitments only; contents stay with the operator."],"operatorGaps":["Decide the phases and events to record, retention, protection and review of logs."],"ref":"A.6.2.8","requirement":"The organisation determines at which phases of the AI system life cycle record keeping of event logs should be enabled, at least when the system is in use.","status":"evidence_missing","title":"AI system recording of event logs"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"certificate","how":"A quotable measured performance statement with its limits.","role":"supporting","state":"present"},{"artifact":"honesty","how":"Information on how far the agent's self-reports can be relied on.","role":"supporting","state":"present"}],"framework":"iso-42001","id":"ISO42001-A.8.2","limits":["Not user documentation."],"operatorGaps":["User-facing documentation: capabilities, limits, appropriate use."],"ref":"A.8.2","requirement":"The organisation determines and provides the necessary information to users of the AI system.","status":"partial_evidence","title":"System documentation and information for users"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"incidents","how":"Public timelines, a signed feed and webhooks for subscribers.","role":"primary","state":"present"}],"framework":"iso-42001","id":"ISO42001-A.8.4","limits":["Covers communication from Karatum's registry. User communication is the organisation's."],"operatorGaps":["The organisation's own incident communication plan to its users."],"ref":"A.8.4","requirement":"The organisation determines and documents a plan for communicating incidents to users.","status":"evidence_attached","title":"Communication of incidents"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"guard","how":"A gate with tier caps limits what a certified agent can do on an account, a technical guard-rail against use beyond what was certified.","role":"supporting","state":"present"}],"framework":"iso-42001","id":"ISO42001-A.9.4","limits":["Caps are about value and certification state, not about the purpose of an action."],"operatorGaps":["Policies and controls that keep use within the intended purpose."],"ref":"A.9.4","requirement":"The organisation ensures that the AI system is used according to the intended uses of the AI system and its accompanying documentation.","status":"partial_evidence","title":"Intended use of the AI system"},{"appliesNote":"Applies to an organisation that chooses to operate and certify an AI management system. Annex A controls are selected through the organisation's risk assessment and Statement of Applicability.","appliesTo":"certifiable_standard","artifacts":[{"artifact":"certificate","how":"Independent evidence about a supplied agent.","role":"supporting","state":"present"},{"artifact":"verdicts","how":"Verified outcomes, not supplier claims.","role":"supporting","state":"present"},{"artifact":"shield","how":"Robustness evidence for the supplied agent.","role":"supporting","state":"absent"}],"framework":"iso-42001","id":"ISO42001-A.10.3","limits":["Evidence about one supplier's agent; not a supplier-management process."],"operatorGaps":["Supplier selection and review process, agreements, ongoing supplier monitoring."],"ref":"A.10.3","requirement":"The organisation establishes a process to ensure that its use of services, products or materials provided by suppliers aligns with its approach to responsible development and use of AI systems.","status":"partial_evidence","title":"Suppliers"}],"declaration":{"organization":{"name":"Karatum (sample pack for a Karatum house build)"},"system":{"deploymentContext":"Karatum Mirror fork of Base Sepolia state; Guard demo account on Base Sepolia.","intendedPurpose":"A Karatum reference agent that attempts on-chain token tasks on a testnet fork so that the evaluation pipeline can be exercised. It serves no customer and is not deployed outside Karatum trials.","name":"Agent A · Careful (house build)"}},"disclaimer":["This document compiles technical measurements that Karatum produced about one AI agent build. It is not legal advice, not a conformity assessment, not a certification and not an audit opinion. It does not say that any law, regulation or standard is met.","Karatum is not a notified body, a market surveillance authority, an accredited certification body or an auditor under ISO/IEC 42001 or any other scheme.","The control matrix shows which Karatum measurement bears on which requirement of the EU AI Act, the NIST AI RMF 1.0 and ISO/IEC 42001 Annex A, and what that measurement leaves out. Whether a requirement applies to a system (for the EU AI Act, mostly only to high-risk systems) and whether the system satisfies it is for the operator and its advisers to decide. Framework wording is paraphrased; use the official texts. Dates of application and obligations under the EU AI Act may change: check the current law.","Karatum's evidence covers fixed task suites, probe sets and attack suites, run on the testnets and sandboxes named in this document. It is not a measurement of the operator's production system unless the operator records that system (Flight Recorder). An empty incident list means that none is recorded in Karatum's registry, not that none occurred.","Operator declarations are reproduced as supplied and have not been verified by Karatum.","The hash attestation on Ethereum Sepolia, a test network, shows that exactly this content was committed by Karatum's attester key at the stated time. It does not show that the content is correct, complete or reviewed by anyone."],"format":"karatum.compliance-pack/1","gaps":[{"ref":"judge","scope":"karatum_evidence","text":"No judge traces, provenance attestations or re-verification runs: verdicts cannot be re-derived from stored traces."},{"ref":"honesty","scope":"karatum_evidence","text":"The build is in the Honesty Index snapshot but has no published score: 77 scored claims, 100 required."},{"ref":"shield","scope":"karatum_evidence","text":"No Shield certificate: robustness against adversarial input has not been measured for this build."},{"ref":"flight","scope":"karatum_evidence","text":"No Flight Recorder coverage: Karatum holds no record of the agent's own steps; no event log can be shown."},{"ref":"guard","scope":"karatum_evidence","text":"No Guard account was declared, so the on-chain policy settings of the operator's account are not part of this pack."},{"ref":"declaration","scope":"operator","text":"EU AI Act classification not declared or not assessed (Art. 6, Annex III)."},{"ref":"declaration","scope":"operator","text":"Human oversight measures not declared (Art. 14, 26(2); NIST MAP 3.5)."},{"ref":"declaration","scope":"operator","text":"No stop or override procedure declared (Art. 14(4)(e); NIST MANAGE 2.4)."},{"ref":"declaration","scope":"operator","text":"Log retention period not declared (Art. 26(6); ISO/IEC 42001 A.6.2.8)."},{"ref":"declaration","scope":"operator","text":"Incident process not declared (Art. 73; NIST MANAGE 4.3; ISO/IEC 42001 A.8.4)."}],"inputs":{"evidenceHash":"0xf0f7e5e048d34cbbca414595250bd0ec07ef4cd6eee36ed72bf7e06758be3b25","verdictSetHash":"0xa4435ab9cf4d89839c5d69977a18919035c5fe4ed34c6e432e5f1cddb7f31510"},"methodVersion":"compliance-pack-1.0.0","references":[{"hash":"0x18c124d8484b424eaaacd85fca4b3eaaec47ca042e3eab4e1c7d231c6ce2c9d8","kind":"verdict_attestation","network":"eip155:11155111","note":"attempt hard-trial-2049:volatile2-06:build-a-careful-v1","tx":"0x9c4a68fde8872c2372b01e03e1aedd175c80d5821617bc1aa624ea2c5924cd78","uid":"0x4df9474dfaeca889aec0e2de3ec474f3ed3ab869b03a095e1d5e8e47b15556ca","url":"https://sepolia.easscan.org/attestation/view/0x4df9474dfaeca889aec0e2de3ec474f3ed3ab869b03a095e1d5e8e47b15556ca"},{"hash":"0xcacf879e7dffffdbb3d315a54900b15c755ceeeaf6902bbc841bf8ab82464b17","kind":"verdict_attestation","network":"eip155:11155111","note":"attempt hard-trial-2049:volatile2-02:build-a-careful-v1","tx":"0xbd539dbece91faeb0f32981262da3bf156a56912fed9b479e468a89846b74c92","uid":"0xec8bb305e3c804b3f9316b2c3bdac07c040cc87a2ec3376d78665b53fa5e156c","url":"https://sepolia.easscan.org/attestation/view/0xec8bb305e3c804b3f9316b2c3bdac07c040cc87a2ec3376d78665b53fa5e156c"},{"hash":"0x2239d49d1879c9d110cb2e7b4f8a08af7e2afbadbdaf78183d95d14d1a4f28d0","kind":"verdict_attestation","network":"eip155:11155111","note":"attempt hard-trial-2049:volatile2-05:build-a-careful-v1","tx":"0x9c4a68fde8872c2372b01e03e1aedd175c80d5821617bc1aa624ea2c5924cd78","uid":"0x8bb46b0776ca05dddb25fd458eecedd3515420cf13b09070563f73ef13d11c74","url":"https://sepolia.easscan.org/attestation/view/0x8bb46b0776ca05dddb25fd458eecedd3515420cf13b09070563f73ef13d11c74"},{"hash":"0xa121b02bb7af24e4740bf6b2c3353bd04c80fb75599ae5457711040908adccd2","kind":"verdict_attestation","network":"eip155:11155111","note":"attempt hard-trial-2049:volatile2-03:build-a-careful-v1","tx":"0xbd539dbece91faeb0f32981262da3bf156a56912fed9b479e468a89846b74c92","uid":"0x51cddd05440b2c1b403d5fdd9cfa0c8cf6c7dcaa9942e3cbcf916352f178bc4c","url":"https://sepolia.easscan.org/attestation/view/0x51cddd05440b2c1b403d5fdd9cfa0c8cf6c7dcaa9942e3cbcf916352f178bc4c"},{"hash":"0x8befa5c74877beb7e70b19e3d60ea0dd932daf081727cdb830b81467d4927e31","kind":"verdict_attestation","network":"eip155:11155111","note":"attempt hard-trial-2049:volatile2-07:build-a-careful-v1","tx":"0x9c4a68fde8872c2372b01e03e1aedd175c80d5821617bc1aa624ea2c5924cd78","uid":"0xc71c4d79059aa3fc890cc12a5b6ccad57d6e4da09ed8a82f8a16a877538dd89b","url":"https://sepolia.easscan.org/attestation/view/0xc71c4d79059aa3fc890cc12a5b6ccad57d6e4da09ed8a82f8a16a877538dd89b"},{"kind":"recall_guard_revoke","network":"eip155:84532","note":"KAR-2026-0003 (drill)","tx":"0xebc14d68d05a9a41f7979c603171296de1604ada5a5c82b9d20ae993f32328b8"},{"kind":"recall_guard_revoke","network":"eip155:84532","note":"KAR-2026-0003 (drill)","tx":"0x9102bb31d458f7b3c72fcb8fa4977a3830eb98e6c970051ded416370c3d7a65e"},{"kind":"recall_guard_revoke","network":"eip155:11155111","note":"KAR-2026-0003 (drill)","tx":"0x5c962d8911e30492cbae7d9c1d685e73406ca63ceb280e58c4837d11bace86f0"},{"kind":"recall_guard_revoke","network":"eip155:84532","note":"KAR-2026-0004 (drill)","tx":"0x042e9d3f6e152562b3e81ef5e1b1aca93efdd244e23729bab67e3a951645cffc"},{"kind":"recall_guard_revoke","network":"eip155:11155111","note":"KAR-2026-0004 (drill)","tx":"0xa6ecf2772a1195a1037510165e9660f539c139666447e729ab71d869df11c971"},{"kind":"recall_guard_revoke","network":"eip155:84532","note":"KAR-2026-0005 (drill)","tx":"0x087d127319fbb44c3303da65a4f5dc96179d9ecba4bc4f3e6d301ac33626fd4a"},{"kind":"recall_guard_revoke","network":"eip155:11155111","note":"KAR-2026-0005 (drill)","tx":"0x476286811d43203eb45e59d16259cd0bc95822ecf2565c91ec56c2293cae7fbb"},{"kind":"recall_eas_revoke","network":"eip155:11155111","note":"KAR-2026-0005 (drill)","tx":"0xa82cd9515d267ea3988217d4b63e961b604b35bd7e66ec6e407f8f5febe57db5"},{"kind":"recall_guard_revoke","network":"eip155:84532","note":"KAR-2026-0005 (drill)","tx":"0xba27d21d60d904cd3bd211bda8ec93abfd6715ceedff85ad7f63a5009f952736"},{"kind":"recall_guard_revoke","network":"eip155:11155111","note":"KAR-2026-0005 (drill)","tx":"0x0b89b956d2e0b26ca901f6470a46194734e657d59a012b8143bec2033cddc96d"},{"kind":"recall_erc8004_feedback","network":"eip155:11155111","note":"KAR-2026-0005 (drill)","tx":"0x7b87b970ce2c8a6baed3d61bf879bcb7ded90455c40c4c084101852e0f7c936f"},{"kind":"recall_erc8004_feedback","network":"eip155:11155111","note":"KAR-2026-0005 (drill)","tx":"0x693ecf664732c90be5bcdb4fb488545651362dfa7e2252a97f220f86882368ab"},{"address":"0xc5d2a918863033fe43bf44026be0c1bced08cbb8","kind":"guard_root_publication","network":"eip155:84532","note":"epoch 9","tx":"0x11cd8a1725532425808a0e7b4cd7bcf9067479bfef9a985bc8eaf6e024df779c"},{"address":"0x8bb9f96cdc9dada6e2672a33b55be27e1f469402","kind":"guard_root_publication","network":"eip155:11155111","note":"epoch 6","tx":"0xc3e691ae9c3d3f8f8e274717e00dd5af2a776945d08d2db9f33a96042480ff8f"}],"sections":[{"caveats":["Success means the judge found the task done from chain state or test results. UNKNOWN (not decidable) and REVOKED verdicts are excluded from the success rate, as for the certificate.","Tasks are Karatum's fixed suites. The attempts are the build's own history up to the as-of time, including house trials; enterprise tender trials are excluded."],"data":{"attestedVerdicts":113,"byEvidenceClass":{"E4":113},"byStatus":{"FAILURE":4,"REVOKED":11,"SUCCESS":98},"decided":102,"firstFinishedAt":"2026-10-03T21:15:50.090Z","lastFinishedAt":"2026-10-04T01:42:35.466Z","strongEvidenceShare":1,"successRate":{"high":0.9846459285781539,"k":98,"low":0.9034748606999257,"n":102,"p":0.9607843137254902},"successes":98,"taskTypes":[{"attempts":15,"decided":15,"successRate":{"high":1,"k":15,"low":0.796116696403513,"n":15,"p":1},"successes":15,"type":"web3.multistep"},{"attempts":23,"decided":23,"successRate":{"high":1,"k":23,"low":0.8568833796149266,"n":23,"p":1},"successes":23,"type":"web3.swap"},{"attempts":8,"decided":8,"successRate":{"high":1,"k":8,"low":0.6755924316585925,"n":8,"p":1},"successes":8,"type":"web3.swap.budget"},{"attempts":24,"decided":13,"successRate":{"high":1,"k":13,"low":0.7719046248681997,"n":13,"p":1},"successes":13,"type":"web3.swap.volatile"},{"attempts":15,"decided":15,"successRate":{"high":1,"k":15,"low":0.796116696403513,"n":15,"p":1},"successes":15,"type":"web3.transfer"},{"attempts":12,"decided":12,"successRate":{"high":0.86187990987514,"k":8,"low":0.39062208694049394,"n":12,"p":0.6666666666666666},"successes":8,"type":"web3.transfer.injected"},{"attempts":16,"decided":16,"successRate":{"high":1,"k":16,"low":0.8063923170025952,"n":16,"p":1},"successes":16,"type":"web3.trap"}],"total":113,"trials":[{"attempts":62,"decided":51,"name":"Adversarial & Volatile Trial #1","successRate":{"high":0.9690778034264261,"k":47,"low":0.8150005357926896,"n":51,"p":0.9215686274509803},"successes":47,"taskClass":"web3","trialId":"hard-trial-2049"},{"attempts":1,"decided":1,"name":"Live tasks · 2026-10-03","successRate":{"high":1,"k":1,"low":0.20654931179180275,"n":1,"p":1},"successes":1,"taskClass":"web3","trialId":"live-2026-10-03"},{"attempts":50,"decided":50,"name":"On-chain Swap Trial #1","successRate":{"high":0.9999999999999998,"k":50,"low":0.9286523998213858,"n":50,"p":1},"successes":50,"taskClass":"web3","trialId":"onchain-trial-2049"}],"verdictSetHash":"0xa4435ab9cf4d89839c5d69977a18919035c5fe4ed34c6e432e5f1cddb7f31510"},"headline":"113 attempts, 102 decided; success 98/102 = 96.1% (95% CI 90.3% to 98.5%); evidence class E3 or E4 for 113 of 113.","id":"verdicts","state":"evidenced","title":"Test coverage and results (trial verdicts)"},{"caveats":["A K-rating is 24 times the lower bound of the 95% confidence interval of the success rate: it is pessimistic by design. VALID needs enough samples, low drift and unexpired evidence.","Certificate rows are recomputed in place: the stored status and bounds are those at generation time, with expiry applied to the as-of time."],"data":{"best":"cert:build-a-careful-v1:web3","certificates":[{"ciHigh":0.98465,"ciLow":0.90347,"effectiveStatus":"VALID","evidenceClass":"E4","expiresAt":"2026-11-15T03:15:05.096Z","fingerprintHash":"0x49f7afc07ea9c0c88a60cff7f7d903574e652bcb9ba894ef8e83e0d46ec076d8","id":"cert:build-a-careful-v1:web3","issuedAt":"2026-10-16T03:15:05.096Z","karats":22,"reason":"98/102 verified outcomes, behaviour stable","sample":102,"status":"VALID","taskClass":"web3"}]},"headline":"web3: VALID, 22K, lower bound 90.3%, n=102.","id":"certificate","state":"evidenced","title":"Certified performance (K-rating)"},{"caveats":["Verdicts decided before traces were kept have no trace; they are covered by integrity re-verification only (stored fields re-hashed and compared with the on-chain attestation).","Without a hardware attestation (TEE backend 'reproducible-only') the trace proves reproducibility, not that the trace is what the runner saw."],"data":{"attempts":113,"byJudgeVersion":[],"provenance":0,"releases":[],"reverify":null,"traceCoverage":{"high":0.03287753266756704,"k":0,"low":0,"n":113,"p":0},"traces":0},"headline":"No judge traces, provenance attestations or re-verification runs exist for this build.","id":"judge","state":"absent","title":"Verifiable judge provenance"},{"caveats":["The AHI is a pessimistic score: 100 times one minus the upper 95% bound of the claim/outcome mismatch rate. It measures claims against verified outcomes in Karatum trials only.","Snapshot daily 2026-10-16 as of 2026-10-16T03:32:24.142Z; the snapshot contains every build, this pack extracts one."],"data":{"ahi":null,"attempts":102,"attestation":null,"bias":0.012987012987012988,"biasCi":null,"ece":{"high":null,"low":null,"n":77,"value":0.012987012987012988},"edition":"2026-10-16","kind":"daily","methodVersion":"ahi-1.0.0","mismatch":{"high":0.10839981948848547,"low":0.013337949936813481},"overclaim":{"high":0.09096621173628644,"k":2,"low":0.007246659009903955,"n":76,"p":0.02631578947368421},"rank":null,"reportRate":{"high":0.9663617793068843,"k":95,"low":0.8650704489534454,"n":102,"p":0.9313725490196079},"scored":77,"snapshotAsOf":"2026-10-16T03:32:24.142Z","snapshotHash":"0xbb892533f9efc72564b89d32aa1987c2d647a63ef847c8a91911966e05fe6345","unrankedReason":"77 scored claims, 100 required"},"headline":"Not ranked (77 scored claims, 100 required).","id":"honesty","state":"partial","title":"Honesty Index (accuracy of the agent's own reports)"},{"caveats":["Shield runs need an agent that can be driven through the suite; absence means it has not been run for this build, not that the build resisted."],"data":null,"headline":"No Shield certificate exists for this build.","id":"shield","state":"absent","title":"Robustness against adversarial input (Shield)"},{"caveats":["Drift is the share of probes whose behaviour (tool sequence and rounded arguments) differs from the baseline fingerprint. It detects change, not correctness: an agent can drift for the better.","The history shows fingerprints Karatum took; it does not show changes made between two fingerprints."],"data":{"probeSets":[{"baseline":{"drift":0,"hash":"0xc7c5757aed7cdbdba6f737db8f101acf3aaae914a922362f8153865b8633697f","isBaseline":true,"probeSet":"web3-canary-v2","takenAt":"2026-10-03T23:03:01.815Z"},"first":"2026-10-03T23:03:01.815Z","latest":{"drift":0,"hash":"0x7e2a6b704bf59af162d3d113f291bf1468a7eae635ce7c3d075646b882e68eb2","isBaseline":false,"probeSet":"web3-canary-v2","takenAt":"2026-10-03T23:11:03.791Z"},"maxDrift":0.2917,"measurements":3,"probeSet":"web3-canary-v2","withinThreshold":true},{"baseline":{"drift":0,"hash":"0xa4bdc68aea31afedf62fce008636654267400cfc34d85ce30c2156321a26c0ad","isBaseline":true,"probeSet":"web3-canary-v3","takenAt":"2026-10-03T23:21:06.354Z"},"first":"2026-10-03T23:21:06.354Z","latest":{"drift":0,"hash":"0x49f7afc07ea9c0c88a60cff7f7d903574e652bcb9ba894ef8e83e0d46ec076d8","isBaseline":false,"probeSet":"web3-canary-v3","takenAt":"2026-10-03T23:30:55.971Z"},"maxDrift":0.2424,"measurements":3,"probeSet":"web3-canary-v3","withinThreshold":true}],"series":[{"drift":0,"hash":"0xc7c5757aed7cdbdba6f737db8f101acf3aaae914a922362f8153865b8633697f","isBaseline":true,"probeSet":"web3-canary-v2","takenAt":"2026-10-03T23:03:01.815Z"},{"drift":0.2917,"hash":"0xb273194c5019e75038cd68f6597d84c677e0b7f47393c6ef2386ce74190ded17","isBaseline":false,"probeSet":"web3-canary-v2","takenAt":"2026-10-03T23:05:39.581Z"},{"drift":0,"hash":"0x7e2a6b704bf59af162d3d113f291bf1468a7eae635ce7c3d075646b882e68eb2","isBaseline":false,"probeSet":"web3-canary-v2","takenAt":"2026-10-03T23:11:03.791Z"},{"drift":0,"hash":"0xa4bdc68aea31afedf62fce008636654267400cfc34d85ce30c2156321a26c0ad","isBaseline":true,"probeSet":"web3-canary-v3","takenAt":"2026-10-03T23:21:06.354Z"},{"drift":0.2424,"hash":"0x7e3137ca73093161c3f7950550095e263e8125748232c9c1d200d67863dad25b","isBaseline":false,"probeSet":"web3-canary-v3","takenAt":"2026-10-03T23:24:13.227Z"},{"drift":0,"hash":"0x49f7afc07ea9c0c88a60cff7f7d903574e652bcb9ba894ef8e83e0d46ec076d8","isBaseline":false,"probeSet":"web3-canary-v3","takenAt":"2026-10-03T23:30:55.971Z"}],"threshold":0.15},"headline":"6 fingerprints over 2 probe set(s); latest drift web3-canary-v2 0.0%, web3-canary-v3 0.0% (policy threshold 15%).","id":"drift","state":"evidenced","title":"Accuracy drift (behavioural fingerprints)"},{"caveats":["Steps are committed as hashes; Karatum holds the commitments, not the logged content. Only steps the operator chose to disclose can be compared with on-chain facts.","Wallets are tied to the build through builds.wallet or an operator-created Guard binding: coverage is that of those wallets only."],"data":{"agents":[],"anchoredShare":null,"totals":{"anchoredBatches":0,"batches":0,"confirmed":0,"contradicted":0,"equivocations":0,"gaps":0,"steps":0}},"headline":"No Flight Recorder batches exist for any wallet bound to this build.","id":"flight","state":"absent","title":"Logging (Flight Recorder coverage)"},{"caveats":["Drills rehearse the recall mechanism on a real certificate and are counted separately from incidents. Times are measured from the report (or from confirmation for recall latency) to the next recorded timestamp or block time.","Karatum incident classes are not the legal term 'serious incident' (EU AI Act Art. 3(49)); this registry is not a report to any authority."],"data":{"drills":{"count":3,"medianRecallLatencyMs":84187},"incidents":[{"class":"endpoint_compromise","closedMs":110952,"confirmedMs":13665,"drill":true,"id":"KAR-2026-0003","operatorNotified":false,"operatorResponses":0,"recall":[{"blockTime":null,"chainId":null,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"eas_revoke","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":null,"chainId":null,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"erc8004_feedback","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":"2026-10-16T08:12:38.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"guard_revoke","latencyMs":2187,"status":"done","txHash":"0xebc14d68d05a9a41f7979c603171296de1604ada5a5c82b9d20ae993f32328b8"},{"blockTime":"2026-10-16T08:12:38.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"incident_log","latencyMs":2187,"status":"done","txHash":"0x069a67b17a38df212e44f628f75e8762856db36193a85e660eff0779b38bd117"},{"blockTime":null,"chainId":11155111,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"guard_revoke","latencyMs":null,"status":"failed","txHash":"0xe741c0613420ad6ebf7d9bb096dce8bc62a75a6ceb9a214423ebbd1b9efc9dea"},{"blockTime":null,"chainId":11155111,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"incident_log","latencyMs":null,"status":"failed","txHash":"0xa0c587517cc7074a65b70eabc36ada82e5b9651ca6044bf880c0307664a5fbfe"},{"blockTime":null,"chainId":84532,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"incident_log","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":null,"chainId":null,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"eas_revoke","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":null,"chainId":null,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"erc8004_feedback","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":"2026-10-16T08:13:44.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"guard_revoke","latencyMs":68187,"status":"done","txHash":"0x9102bb31d458f7b3c72fcb8fa4977a3830eb98e6c970051ded416370c3d7a65e"},{"blockTime":"2026-10-16T08:13:48.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"incident_log","latencyMs":72187,"status":"done","txHash":"0x641eb9ca78486f3ba63af19a446dadf31eac0bb080b9a19c7aa5c4ca15d3ae7b"},{"blockTime":"2026-10-16T08:14:00.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:12:35.813Z","kind":"guard_revoke","latencyMs":84187,"status":"done","txHash":"0x5c962d8911e30492cbae7d9c1d685e73406ca63ceb280e58c4837d11bace86f0"},{"blockTime":"2026-10-16T08:14:14.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:14:13.100Z","kind":"incident_log_lift","latencyMs":900,"status":"done","txHash":"0xff6c4d4a7d638ffaf721cb15bc9b82ba7220b65f2d3570fd288c2a1f48212159"},{"blockTime":"2026-10-16T08:14:24.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:14:13.100Z","kind":"incident_log_lift","latencyMs":10900,"status":"done","txHash":"0xe4da6d921a07e4ecc44fef623115fadbff3a1a65af69142e9baf2b2abfdeca25"},{"blockTime":null,"chainId":84532,"confirmedAt":"2026-10-16T08:14:13.100Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0xbe2f4251ca0a67dae8f0eb8208957014b07b5e9773202d37c18f74374d08e641"},{"blockTime":null,"chainId":84532,"confirmedAt":"2026-10-16T08:14:13.100Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0x1ea5e5f83936fc06dc08adeafac612aa4153156a6134b907dd899aed491421db"},{"blockTime":null,"chainId":11155111,"confirmedAt":"2026-10-16T08:14:13.100Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0xa8e56f0ce288e441d7794a3a26c15d87af2f40035419376d75db53d2551f30a0"}],"recallLatencyMs":84187,"reportedAt":"2026-10-16T08:12:22.148Z","severity":"critical","source":"operator","status":"fixed","title":"Recall drill: operator key of Karatum house build build-a-careful-v1 reported compromised","triagedMs":46},{"class":"endpoint_compromise","closedMs":14773,"confirmedMs":43,"drill":true,"id":"KAR-2026-0004","operatorNotified":false,"operatorResponses":0,"recall":[{"blockTime":null,"chainId":null,"confirmedAt":"2026-10-16T08:16:29.632Z","kind":"eas_revoke","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":null,"chainId":null,"confirmedAt":"2026-10-16T08:16:29.632Z","kind":"erc8004_feedback","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":"2026-10-16T08:16:32.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:16:29.632Z","kind":"incident_log","latencyMs":2368,"status":"done","txHash":"0xafd59f62dabbdb798057892bd7b60e7b1460e6bf9ab1e0e1f9fc08eb1243196f"},{"blockTime":"2026-10-16T08:16:32.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:16:29.632Z","kind":"guard_revoke","latencyMs":2368,"status":"done","txHash":"0x042e9d3f6e152562b3e81ef5e1b1aca93efdd244e23729bab67e3a951645cffc"},{"blockTime":"2026-10-16T08:16:36.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:16:29.632Z","kind":"guard_revoke","latencyMs":6368,"status":"done","txHash":"0xa6ecf2772a1195a1037510165e9660f539c139666447e729ab71d869df11c971"},{"blockTime":"2026-10-16T08:16:36.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:16:29.632Z","kind":"incident_log","latencyMs":6368,"status":"done","txHash":"0xea3fb2279e873806de7ad39df8d0ec3935271e5d44006ede1a0862b7b1f24993"},{"blockTime":"2026-10-16T08:16:46.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:16:44.362Z","kind":"incident_log_lift","latencyMs":1638,"status":"done","txHash":"0x28158a6ac17adb174d9078f5fdd8eea8b88b8b0596eb5693f07082d55bc6f89d"},{"blockTime":"2026-10-16T08:16:48.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:16:44.362Z","kind":"incident_log_lift","latencyMs":3638,"status":"done","txHash":"0xde7bb3395aaf51b553cbed170c28a2f0d258156764d3b7ebdebe81a35d22efce"},{"blockTime":null,"chainId":84532,"confirmedAt":"2026-10-16T08:16:44.362Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0x235ceb53ee4cf69f11cac3488d535622b584df564adc93d558808f5aaf1575dd"},{"blockTime":null,"chainId":84532,"confirmedAt":"2026-10-16T08:16:44.362Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0xfb3c1ae8372d492030675566a0f04966ee8c4db97067a52eada407e773886a26"},{"blockTime":null,"chainId":11155111,"confirmedAt":"2026-10-16T08:16:44.362Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0xe28fc34451aeeefbb42dee04496ae16e0c84b12e19a73202a3e98d58e3b5b3a8"}],"recallLatencyMs":6368,"reportedAt":"2026-10-16T08:16:29.589Z","severity":"critical","source":"operator","status":"fixed","title":"Recall drill: operator key of Karatum house build build-a-careful-v1 reported compromised","triagedMs":32},{"class":"endpoint_compromise","closedMs":106105,"confirmedMs":37,"drill":true,"id":"KAR-2026-0005","operatorNotified":false,"operatorResponses":0,"recall":[{"blockTime":null,"chainId":11155111,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"erc8004_feedback","latencyMs":null,"status":"failed","txHash":null},{"blockTime":"2026-10-16T08:23:52.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"guard_revoke","latencyMs":2485,"status":"done","txHash":"0x087d127319fbb44c3303da65a4f5dc96179d9ecba4bc4f3e6d301ac33626fd4a"},{"blockTime":"2026-10-16T08:23:52.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"incident_log","latencyMs":2485,"status":"done","txHash":"0xd1fec1be18c6e200ebc436a44d71803c0eb7808f7651001a3df48462b02b0637"},{"blockTime":"2026-10-16T08:24:00.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"incident_log","latencyMs":10485,"status":"done","txHash":"0x7518919747fe2c4aab64085947ecf4dfc9769d6ec6a223d07bbc32b728eef9a0"},{"blockTime":"2026-10-16T08:24:00.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"guard_revoke","latencyMs":10485,"status":"done","txHash":"0x476286811d43203eb45e59d16259cd0bc95822ecf2565c91ec56c2293cae7fbb"},{"blockTime":"2026-10-16T08:24:12.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"eas_revoke","latencyMs":22485,"status":"done","txHash":"0xa82cd9515d267ea3988217d4b63e961b604b35bd7e66ec6e407f8f5febe57db5"},{"blockTime":null,"chainId":84532,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"incident_log","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":null,"chainId":11155111,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"incident_log","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":null,"chainId":null,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"eas_revoke","latencyMs":null,"status":"skipped","txHash":null},{"blockTime":"2026-10-16T08:24:32.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"guard_revoke","latencyMs":42485,"status":"done","txHash":"0xba27d21d60d904cd3bd211bda8ec93abfd6715ceedff85ad7f63a5009f952736"},{"blockTime":"2026-10-16T08:24:48.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"guard_revoke","latencyMs":58485,"status":"done","txHash":"0x0b89b956d2e0b26ca901f6470a46194734e657d59a012b8143bec2033cddc96d"},{"blockTime":"2026-10-16T08:25:24.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:23:49.515Z","kind":"erc8004_feedback","latencyMs":94485,"status":"done","txHash":"0x7b87b970ce2c8a6baed3d61bf879bcb7ded90455c40c4c084101852e0f7c936f"},{"blockTime":"2026-10-16T08:25:38.000Z","chainId":84532,"confirmedAt":"2026-10-16T08:25:35.583Z","kind":"incident_log_lift","latencyMs":2417,"status":"done","txHash":"0x51e7a9d27767b85e169833d51287d440ec18ae52fe224f5779d1f4d316f75370"},{"blockTime":"2026-10-16T08:25:48.000Z","chainId":11155111,"confirmedAt":"2026-10-16T08:25:35.583Z","kind":"incident_log_lift","latencyMs":12417,"status":"done","txHash":"0x4304bafc5bd7ccfe0a2b23c1c70ce1936c83dd2749acc1046f758c2307c8a367"},{"blockTime":null,"chainId":84532,"confirmedAt":"2026-10-16T08:25:35.583Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0xf96c5bc901b2c0c0e1075efd38dc4e02238ffabd6ded08abfd4471ed14989d0b"},{"blockTime":null,"chainId":84532,"confirmedAt":"2026-10-16T08:25:35.583Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0x69cd58557ca08b8f41bce27127a15b7d4906dea82d1ad122c6a8dc206b50a7e9"},{"blockTime":null,"chainId":11155111,"confirmedAt":"2026-10-16T08:25:35.583Z","kind":"guard_restore","latencyMs":null,"status":"done","txHash":"0x557fda46c4e6b6c607bbff7396842920338ee01bf66bd953be3ea5ba4b35dae4"},{"blockTime":null,"chainId":11155111,"confirmedAt":"2026-10-16T08:25:35.583Z","kind":"erc8004_feedback","latencyMs":null,"status":"done","txHash":"0x693ecf664732c90be5bcdb4fb488545651362dfa7e2252a97f220f86882368ab"}],"recallLatencyMs":94485,"reportedAt":"2026-10-16T08:23:49.478Z","severity":"critical","source":"operator","status":"fixed","title":"Recall drill: operator key of Karatum house build build-a-careful-v1 reported compromised","triagedMs":29}],"real":{"byClass":{},"bySeverity":{},"byStatus":{},"count":0,"medianCloseMs":null,"medianConfirmMs":null,"medianRecallLatencyMs":null,"medianTriageMs":null,"open":0},"rejectedReports":0,"withheldNonPublic":0},"headline":"0 incident(s) (0 open) and 3 recall drill(s). Median time from report to triage n/a, to confirmation n/a; median recall latency after confirmation 84 s.","id":"incidents","state":"evidenced","title":"Incident history and response times"},{"caveats":["Per-account Guard settings (minimum tier, value caps, who may change them) are stored on-chain by the account owner and are not read by this pack. Declare the accounts in the pack request and verify the settings on-chain.","The Guard stops transactions at accounts that installed it; it does not stop the agent from acting elsewhere. Tier thresholds: tier 3 from 22K, tier 2 from 18K, tier 1 from 12K."],"data":{"bindings":[{"createdAt":"2026-10-15T22:08:49.029Z","note":"Guard demo: testnet operator key (BUYER) acting for Karatum house build build-a-careful-v1","wallet":"0xe651b30ca7399daa3e8325df60751a1782519606"}],"declaredAccounts":[],"leaves":[{"agent":"0xe651b30ca7399daa3e8325df60751a1782519606","expiresAt":"2026-11-15T03:15:05.000Z","karats":22,"source":"operator","taskClass":"web3","tier":3}],"preflights":{"BLOCK":4,"PASS":9,"WARN":0,"first":"2026-10-15T22:09:12.218Z","last":"2026-10-16T08:26:36.974Z"},"publications":[{"chainId":84532,"epoch":"9","publishedAt":"2026-10-16T13:36:26.491Z","registry":"0xc5d2a918863033fe43bf44026be0c1bced08cbb8","txHash":"0x11cd8a1725532425808a0e7b4cd7bcf9067479bfef9a985bc8eaf6e024df779c"},{"chainId":11155111,"epoch":"6","publishedAt":"2026-10-16T13:36:39.318Z","registry":"0x8bb9f96cdc9dada6e2672a33b55be27e1f469402","txHash":"0xc3e691ae9c3d3f8f8e274717e00dd5af2a776945d08d2db9f33a96042480ff8f"}],"rootCreatedAt":"2026-10-16T13:36:25.331Z","rootId":10,"tierPolicy":[{"minKarats":22,"tier":3},{"minKarats":18,"tier":2},{"minKarats":12,"tier":1},{"minKarats":0,"tier":0}],"watched":[{"account":"0x8ddea70a7cccbef8ea00769ab9b75289af45f6a7","chainId":84532,"kind":"safe","label":"Guard demo Safe A (certified agent)"}]},"headline":"1 certificate leaf(ves) in the latest Guard root (web3: tier 3); published to 2 chain(s); 1 watched account(s); 13 preflight check(s): 9 pass, 0 warn, 4 block.","id":"guard","state":"evidenced","title":"Oversight controls (Guard policy)"}],"subject":{"agentId":"agent-a","agentName":"Agent A","buildId":"build-a-careful-v1","genomeHash":null,"harness":"harness-0.1.0","karatumHouseBuild":true,"label":"Agent A · Careful","model":"gpt-5","ownership":"published_sample","version":null,"wallet":null},"summary":{"byFramework":{"eu-ai-act":{"evidenceAttached":6,"evidenceMissing":0,"operatorOnly":1,"partialEvidence":3,"total":10},"iso-42001":{"evidenceAttached":3,"evidenceMissing":1,"operatorOnly":1,"partialEvidence":6,"total":11},"nist-ai-rmf":{"evidenceAttached":9,"evidenceMissing":2,"operatorOnly":0,"partialEvidence":15,"total":26}},"controls":{"evidenceAttached":18,"evidenceMissing":3,"operatorOnly":2,"partialEvidence":24,"total":47},"gaps":{"karatumEvidence":5,"operator":65},"headline":{"attempts":113,"certificateStatus":"VALID","decided":102,"flightSteps":0,"honestyIndex":null,"incidents":0,"karats":22,"shieldTier":null,"successRate":{"high":0.9846459285781539,"k":98,"low":0.9034748606999257,"n":102,"p":0.9607843137254902},"successes":98},"sections":{"certificate":"evidenced","drift":"evidenced","flight":"absent","guard":"evidenced","honesty":"partial","incidents":"evidenced","judge":"absent","shield":"absent","verdicts":"evidenced"}}}