Examples
Examples from published documents
- During UK AISI testing an agent's malicious pull request to an open-source project "was caught by a human reviewer"; AISI noted the margin "was narrow, resting on human vigilance rather than a technical barrier" (UK AI Security Institute, Incident Report, 4 Aug 2026)
Counter-examples
Not to be confused with
- A near miss is not an incident with low harm: under the EU definition the serious incident did not materialise at all, while a low-severity incident is one where some harm did occur.
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.
- AI hazard
- AI hazard is the OECD's term for an event, circumstance or series of events where the development, use or malfunction of AI systems could plausibly lead to an AI incident. The OECD counts near misses as AI hazards.
- Source: OECD, Defining AI incidents and related terms (OECD Artificial Intelligence Papers No. 16, 2024)
- Serious AI hazard
- Serious AI hazard is the OECD's term for an AI hazard that could plausibly lead to a serious AI incident or an AI disaster, such as a death, serious harm to health or a serious and irreversible disruption of critical infrastructure.
- Source: OECD, Defining AI incidents and related terms (OECD Artificial Intelligence Papers No. 16, 2024)
- Precursor event
- Precursor event is the Frontier Model Forum's term for a lower-risk event, such as an anomaly, near-miss, false positive or concerning signal, that falls below formal reporting thresholds but may still be safety-relevant.
- Source: Frontier Model Forum -- Information Sharing Issue Brief
Source of record
Where this definition comes from
EU GPAI Code of Practice, Safety and Security Chapter, Glossary
“"'near miss': a situation in which a serious incident could have, but ultimately did not, materialise."”
https://ec.europa.eu/newsroom/dae/redirection/document/118119OECD, Defining AI incidents and related terms (OECD Artificial Intelligence Papers No. 16, 2024), Draft definition of an AI hazard, key clarifications
“"Near misses are events that could have led to an AI incident and are therefore included under this definition of AI hazards."”
https://www.oecd.org/content/dam/oecd/en/publications/reports/2024/05/defining-ai-incidents-and-related-terms_88d089ec/d1a8d965-en.pdfFrontier Model Forum -- Information Sharing Issue Brief, Incident Reporting
“"Not every anomaly, near-miss, false positive or concerning signal will meet the threshold for formal incident reporting. These lower-risk incidents or precursor events may nonetheless be safety-relevant, and voluntary information-sharing channels are generally better suited to surfacing" them.”
https://www.frontiermodelforum.org/issue-briefs/information-sharing-incident-reporting-and-incident-response-for-frontier-ai-risks/
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 | “'near miss' (Glossary); reports include "individual or aggregate data on near misses" (Measure 9.2(9)); "serious incidents and/or near misses involving their models or similar models" are grounds for a Framework assessment and a Model Report update.” | exactEQ confidence: high | EU GPAI Code of Practice, Safety and Security Chapter |
| OECDShadow mapping OECD, Defining AI incidents and related terms (2024) | “"Near misses are events that could have led to an AI incident and are therefore included under this definition of AI hazards."” Places near misses inside the broader 'AI hazard' category. | exactEQ confidence: high | OECD, Defining AI incidents and related terms (OECD Artificial Intelligence Papers No. 16, 2024) |
| Frontier Model ForumShadow mapping Frontier Model Forum -- Information Sharing Issue Brief | “"anomaly, near-miss, false positive or concerning signal"; "lower-risk incidents or precursor events" suited to voluntary sharing.” | closeCL 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) | “"The most serious attempts were unsuccessful." "In several cases the margin between failure and success was narrow, resting on human vigilance rather than a technical barrier that would reliably prevent this behaviour in a more capable agent."” AISI reports these as part of an incident, not as separate near misses. | broadBR confidence: medium | 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
Divergence
Where sources materially disagree
The EU Code and the OECD both define near miss by counterfactual harm, but the OECD folds it into 'AI hazard' while the EU keeps it as its own term and makes it a reassessment trigger. The Frontier Model Forum argues near misses belong in voluntary sharing channels rather than mandatory reporting.
Gap
No statute in the corpus requires near misses to be reported individually; the EU Code asks only for patterns and aggregate data within serious-incident reports.
