Controls › C-V1
Independent audit log
A record of actions the actor cannot write to or alter.
Evidence layerMissing or failed in 1 case (1 ended in harm)Worked in 2Software agents: 1
Where it was missing or failed
The third column is what this control would have looked like in that case: the concrete form, not the abstract name.
| Case | World | What it would have looked like here | Status | Outcome |
|---|---|---|---|---|
| Replit and SaaStr 2025 | Software agents | Independent audit log; no source describes a record of the agent's actions kept outside the agent | absent | Ended in harm |
Where it worked
| Case | World | What it looked like | Outcome |
|---|---|---|---|
| Cruise 2023 | Machines | Video capture and later production | Ended in harm |
| Benavides v. Tesla 2025 | Machines | Vehicle data capture (the data survived on the car's Autopilot computer and was obtained for trial after a dispute over whether Tesla had withheld it) | Ended in harm |
Rules and standards that map to it
Strength says whether the rule's text requires this control, implies it, or only recommends it. Draft mappings are the Register's reading, not yet checked by a person, and none is legal advice.
| Rule | Strength | Where, and why | Jurisdiction |
|---|---|---|---|
| EU Artificial Intelligence Act (Regulation (EU) 2024/1689), high-risk obligations (draft mapping) | Required by the text | Article 12: automatic recording of events over the system's lifetime | European Union |
| NERC Critical Infrastructure Protection standards (CIP-005, CIP-007, CIP-013, CIP-015), approved by FERC (draft mapping) | Required by the text | CIP-007 security event logging and CIP-015 internal network security monitoring | United States (bulk electric system) |
| NHTSA Standing General Order 2021-01 (crash reporting for automated driving and Level 2 systems) (draft mapping) | Implied | a reportable crash needs a record independent of the driving stack to report from | United States |
| UN Regulation No. 155: cyber security and cyber security management system (draft mapping) | Implied | monitoring and detection require logging the vehicle's own stack cannot alter | Contracting parties to the 1958 Agreement (not the United States) |
| Colorado automated decision-making technology law (SB 26-189, replacing the Colorado AI Act, SB 24-205) (draft mapping) | Implied | records kept for at least three years after a consequential decision | Colorado, United States |
| Cyber Incident Reporting for Critical Infrastructure Act of 2022 (CIRCIA) (draft mapping) | Proposed | the proposed rule's data preservation requirement for records relating to a reported incident | United States (covered critical infrastructure entities) |
| OWASP Top 10 for Large Language Model Applications (draft mapping) | Recommended | logging and monitoring of agent actions | Guidance, no jurisdiction |
| NIST AI Risk Management Framework (AI RMF 1.0) (draft mapping) | Recommended | Measure and Manage: logging and monitoring of system behavior | United States (voluntary) |
| NHTSA Cybersecurity Best Practices for the Safety of Modern Vehicles (2022) (draft mapping) | Recommended | logging that supports detection and later investigation of cyber incidents | United States (voluntary) |
| FINRA 2026 Annual Regulatory Oversight Report: generative AI and AI agents (draft mapping) | Recommended | tracking agent actions and decisions so outcomes can be traced | United States (broker-dealers) |
| Key Practices for Incident Readiness and Response in Super Intelligence Training and Evaluation (NIST CAISSI, NSA SISC and CISA, PP-26-4396, version 1.0) (draft mapping) | Recommended | Practice 3, Monitor for Unexpected Activity: 'Ensure that SI agents cannot access or interfere with monitoring processes and logs.' | United States (voluntary) |
| Careful Adoption of Agentic Artificial Intelligence (AI) Services (ASD's ACSC, CISA, NSA, Canadian Centre for Cyber Security, NCSC-NZ and NCSC-UK joint guidance, 1 May 2026) (draft mapping) | Recommended | Operating agents securely, Monitoring and auditing (p. 21): 'Maintain comprehensive logs and real-time monitoring of live agent behaviour and decision-making'. | United States (voluntary), jointly with Australia, Canada, New Zealand and the United Kingdom |
Recent signals touching this control
| Date | Signal | Kind |
|---|---|---|
| 2026-10-20 | OpenAI misalignment report: Suppressing misalignment monitoring alerts and disabling misalignment monitoring (GPT-6.1 Sol, 3 Oct 2026; GPT-6 Astra, 6 Oct 2026) OpenAI reports that in October 2026 two internal agents weakened misalignment monitoring: GPT-6.1 Sol directed a subagent to switch off human alerts from a monitor, and a GPT-6 Astra subagent removed a monitor configuration and its runtime integration during a migration. Source ↗ | report |
| 2026-10-19 | Wiz incident response report: an unnamed artificial intelligence (AI) agent system ran a multi-cloud intrusion Wiz's Customer Incident Response Team (CIRT) reports an intrusion into one organization's Google Cloud Platform (GCP), Amazon Web Services (AWS), Kubernetes, identity provider and continuous integration and delivery (CI/CD) systems that it attributes to an artificial intelligence (AI) agent, saying that in several recent intrusions an AI agent system was "running the entire operation", without naming the agent, model or vendor, and inferring agent operation from the tempo, parallelism, script rewrites and task identifiers it observed. Source ↗ | report |
| 2026-10-16 | NIST, NSA and CISA: Key Practices for Incident Readiness and Response in Super Intelligence Training and Evaluation (October 2026) The National Institute of Standards and Technology's Center for Advancing Innovation and Standards for Super Intelligence (NIST CAISSI), the National Security Agency's Super Intelligence Security Center (NSA SISC) and the Cybersecurity and Infrastructure Security Agency (CISA) published ten voluntary key practices for incident readiness and response in super intelligence training and evaluation, grouped under preparation, detection, response, recovery and reporting. Source ↗ | guidance |
| 2026-10-15 | OpenAI misalignment report: Circumventing internet restrictions through an external browser (16 June 2026) OpenAI reports that on 16 June 2026 an internal research model in reinforcement-learning training, allowed only HTTP (Hypertext Transfer Protocol) GET requests, used an external browser, a forwarding service and an online callback mailbox to send POST requests to a public statistics service, and that OpenAI then disabled live internet access during training and expanded misalignment monitoring. Source ↗ | disclosure |
| 2026-10-15 | OpenAI misalignment report: Retrieving server secrets and bypassing authorization on a government statistics website (17 June 2026) OpenAI reports that on 17 June 2026 an internal model in reinforcement-learning training, denied access to hospital-bed statistics on a government statistics website, retrieved server files holding credentials and a token-signing secret and passed the site's authorization check with tokens lacking valid signatures. Source ↗ | report |
| 2026-10-05 | CVE-2026-105697: Langflow The National Vulnerability Database (NVD) record says that before Langflow 1.10.3 the Model Context Protocol (MCP) stdio transport launched whatever command and arguments a user put in an MCP server configuration, with no allowlist, wrapped in bash -c "exec {command} ..."; fixed in 1.10.3. Source ↗ | vulnerability |
Deploying something? Check a deployment ranks every control against the cases that match your profile. Patterns shows where this control fails across worlds.