Examples
Examples from published documents
- Anthropic contained a vendor-access incident "within 90 minutes of learning of it by taking affected systems offline and disabling vendor credentials" and closed the access vector the same day (Anthropic Risk Report, August 2026, §4.5.8.2.1)
- UK AISI terminated all evaluation runs and isolated the relevant machines "within one hour of the alert being raised" (UK AI Security Institute, Incident Report, 4 Aug 2026)
Counter-examples
Not to be confused with
- Status is not severity: a low-severity incident can stay unresolved for months, and a severe one can be contained within an hour.
Related terms
Related AI safety terms
Terms defined under this element that are not NIKOLAI elements themselves, each as the cited document uses it. Every term is also listed in the AI safety glossary.
- Resolved incident
- Resolved incident, under the EU Code of Practice, is a serious incident for which the provider has adopted corrective measures to rectify the harm, if possible, and to assess and mitigate the related systemic risks. Until then the incident is unresolved.
- Source: EU GPAI Code of Practice, Safety and Security Chapter
- Containment
- Containment is the incident-response step of stopping an incident from continuing or spreading, for example by taking affected systems offline, disabling credentials, terminating runs or isolating machines, before longer-term fixes are made.
- Source: Anthropic Risk Report (August 2026)
- Potential AI safety incident
- Potential AI safety incident is OpenAI's term for a detected event not yet assessed. Under its incident response plan it is triaged, investigated, escalated and remediated, then checked against legal criteria for external reporting.
- Source: OpenAI Frontier Governance Framework
Source of record
Where this definition comes from
EU GPAI Code of Practice, Safety and Security Chapter, Glossary
“"'resolved' (serious incident): a serious incident of a model for which the Signatory adopted corrective measures to rectify the harm, if possible, and to assess and mitigate systemic risks related to it. 'Unresolved' is to be understood accordingly."”
https://ec.europa.eu/newsroom/dae/redirection/document/118119OpenAI Frontier Governance Framework, §2.6
“"Potential AI safety incidents are triaged, investigated, escalated, and remediated as appropriate according to the procedures described in the AIRP. These incidents are in turn analyzed to determine whether they meet the criteria for external reporting under applicable laws and regulations."”
https://cdn.openai.com/pdf/e37d949b-8c9f-4d76-b99e-4272f4631a7e/openai-frontier-governance-framework.pdfAnthropic Risk Report (August 2026), §4.5.8.2
“"As any security or safety professional knows, preventing incidents from ever occurring is an unrealistic ideal; this is why it is important to invest in detection, containment, and remediation processes."”
https://www-cdn.anthropic.com/f61d49fa5596956a5dec75fea0e973bf6a6a8378/Redacted%20Risk%20Report%20August%202026%20.pdf
Crosswalk
How named organisations use this concept
| Organisation | Their term, as published | Match & verification | Source |
|---|---|---|---|
| EUShadow mapping EU GPAI Code of Practice, Safety and Security Chapter | “'resolved' / 'Unresolved' (Glossary); "For unresolved serious incidents, Signatories will update the information in their initial report ... in an intermediate report"; final report "not later than 60 days after the serious incident has been resolved" (Measure 9.3).” | exactEQ confidence: high | EU GPAI Code of Practice, Safety and Security Chapter |
| OpenAIShadow mapping OpenAI Frontier Governance Framework | “"Potential AI safety incidents are triaged, investigated, escalated, and remediated as appropriate" (§2.6); the investigation precedes "longer-term solutions to address root causes".” | closeCL confidence: high | OpenAI Frontier Governance Framework |
| AnthropicShadow mapping Anthropic Risk Report (August 2026) | “"detection, containment, and remediation processes" (§4.5.8.2); for one incident: "We contained the activity within 90 minutes of learning of it by taking affected systems offline and disabling vendor credentials. The vulnerable access vector was closed the day we received the report." (§4.5.8.2.1)” | closeCL confidence: medium | Anthropic Risk Report (August 2026) |
| US Congress (S. 5061, 119th Congress, not enacted)Shadow mapping S. 5061, Sec. 4(a)(1) | “Reports may concern "confirmed or suspected artificial intelligence security or artificial intelligence safety incidents" (Sec. 4(a)(1)).” S. 5061 is a bill introduced in the US Senate on 21 July 2026; it has not been enacted. Covers only the suspected/confirmed distinction. | narrowNR confidence: medium | S. 5061 (119th Congress) |
| Frontier Model ForumShadow mapping Frontier Model Forum -- Information Sharing Issue Brief | “Incident response is "the coordinated process of containing, mitigating, and recovering from an incident once one has been identified."” Names the phases; does not define statuses. | broadBR confidence: medium | Frontier Model Forum -- Information Sharing Issue Brief |
| UK AI Security Institute (AISI)Shadow mapping UK AI Security Institute -- Incident Report (4 Aug 2026) | “"We declared a security incident, stopped all related evaluations, and disabled internal access to the most capable models. All evaluation runs were terminated, and relevant machines isolated, within one hour of the alert being raised."” A dated containment record; no status vocabulary. | broadBR confidence: high | UK AI Security Institute -- Incident Report: unsanctioned agent behaviour during cyber testing (4 Aug 2026) |
What do these codes mean?
- exact
- The source term is equivalent to this element
- close
- The source term is close but not equivalent to this element
- broad
- The source term is broader than this element
- narrow
- The source term is narrower than this element
- none
- No mapping claim — used for false-friend and declared-but-undefined rows
- EQ
- Equivalent
- CL
- Close
- BR
- Source is broader than the element
- NR
- Source is narrower than the element
- FF
- False friend — same or similar label, different meaning
- DU
- Declared but undefined by the source
- UV
- Unverified
Related, not mapped
Pointers that are not crosswalk claims
These sources mention this concept but do not define or map it clearly enough to count as a crosswalk row — noted here so the research is visible without overstating it as a mapping.
- OpenAI
Decision rights for misalignment incidents include "pausing or terminating affected activity, isolating systems, imposing containment measures, coordinating affected-party notifications where appropriate, and approving restoration" (§IX.D): restoration is a lifecycle step with an owner.
OpenAI/Hugging Face Incident Technical Report
Divergence
Where sources materially disagree
Only the EU Code of Practice defines a status ('resolved', tied to corrective measures). OpenAI and Anthropic describe the same lifecycle as process steps, and S. 5061 adds the suspected/confirmed distinction at intake.
Gap
No source defines 'closed' or sets a rule for when an incident record is closed, and none publishes time-to-containment or time-to-resolution as standard fields.
