Examples
Examples from published documents
- The EU Code of Practice lets estimates take forms such as "a qualitative systemic risk score (e.g. “moderate” or “critical”)" or "a qualitative systemic risk matrix (e.g. “probability: unlikely” x “impact: high”)" (EU GPAI Code of Practice, Safety and Security Chapter, Measure 3.4)
- Google DeepMind reports that "For each of the four identified domains, we have developed specific scenarios and T/CCLs in which these risks could materialize" (Google DeepMind Frontier Safety Framework v3.1, section 1.3.1)
Counter-examples
Not to be confused with
- The risk acceptance determination comes after this stage and is not part of it: it decides whether the estimated risk is acceptable and what follows (see the NIKOLAI element residual-risk-determination).
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.
- Risk modelling
- Risk modelling, in the EU Code of Practice, is "a structured process aimed at specifying pathways through which a systemic risk stemming from a model might materialise". The Code notes it is "often used interchangeably with the term 'threat modelling'".
- Source: EU GPAI Code of Practice, Safety and Security Chapter
- Risk estimation
- Risk estimation is putting a probability and severity on an identified risk. The EU Code of Practice allows a risk score, risk matrix or probability distribution, for example "a quantitative systemic risk matrix (e.g. 'X-Y%' x 'X-Y EUR damage')".
- Source: EU GPAI Code of Practice, Safety and Security Chapter
- Model-independent information
- Model-independent information, in the EU Code of Practice, is "information, including data and research, that is not tied to a specific model, but can inform systemic risk assessment and mitigation across several models", such as literature reviews and incident databases.
- Source: EU GPAI Code of Practice, Safety and Security Chapter
- Inherent risk assessment
- An inherent risk assessment, in Google DeepMind's FSF v3.1, checks whether a model may, "without appropriate mitigations, contribute to the risks identified", by assessing its capabilities against the critical and tracked capability levels before mitigations are counted.
- Source: Google DeepMind Frontier Safety Framework v3.1
Source of record
Where this definition comes from
EU GPAI Code of Practice, Safety and Security Chapter, Commitments 2 and 3, pp.10-12
“"Systemic risk identification involves two elements: (1) following a structured process to identify the systemic risks stemming from the model (as specified in Measure 2.1); and (2) developing systemic risk scenarios for each identified systemic risk (as specified in Measure 2.2)." "Systemic risk analysis involves five elements for each identified systemic risk, which may overlap and may need to be implemented recursively"”
https://ec.europa.eu/newsroom/dae/redirection/document/118119Google DeepMind Frontier Safety Framework v3.1, s.1.3.1, p.5
“"The first part of our risk management process is risk identification. We identify potential risks that could stem from our models and analyze their characteristics to determine which of the identified risks could be significant or severe risks."”
https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/strengthening-our-frontier-safety-framework/frontier-safety-framework_3-1.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 | “Commitment 2 (identification): "(1) following a structured process to identify the systemic risks stemming from the model ...; and (2) developing systemic risk scenarios for each identified systemic risk". Commitment 3 (analysis): "(1) gathering model-independent information ...; (2) conducting model evaluations ...; (3) modelling the systemic risk ...; and (4) estimating the systemic risk ...; while (5) conducting post-market monitoring".” | exactEQ confidence: high | EU GPAI Code of Practice, Safety and Security Chapter |
| Google DeepMindShadow mapping Google DeepMind Frontier Safety Framework v3.1 | “"The first part of our risk management process is risk identification. We identify potential risks that could stem from our models and analyze their characteristics to determine which of the identified risks could be significant or severe risks." (s.1.3.1), followed by "Inherent Risk Assessment" against the T/CCLs (s.1.3.2).” | exactEQ confidence: high | Google DeepMind Frontier Safety Framework v3.1 |
| xAIShadow mapping xAI Frontier AI Framework (30 Jun 2026, draft-marked) | “Risk identification: "(a) compiling a taxonomy of risks involved in the model's training or deployment, (b) analysing relevant characteristics of these risks, (c) identifying systemic risks stemming from the model and (d) developing systemic risk scenarios and mitigations for each identified systemic risk" (s.2.1); systemic risk modelling "building on the systemic risk scenarios in the risk identification process" (s.2.2(3)).” PDF metadata /Title reads "Privileged/Confidential DRAFT working FRAMEWORK DOC"; no xAI statement disambiguating draft vs. final was found. As of 13 Nov 2026 the cited URL returns an S3 "NoSuchKey" error; the quote is verified against the local capture made on 16 Sep 2026. | exactEQ confidence: medium | xAI Frontier AI Framework (30 Jun 2026, draft-marked) |
| MetaShadow mapping Meta Advanced AI Scaling Framework v2 | “Anticipate stage: "Host workshops with experts, including external subject matter experts where relevant, to identify new catastrophic outcomes and/or threat scenarios" (§2.1.1); "Our AI risk assessment process systematically evaluates potential risks associated with ongoing development and deployment of Frontier AI" (§2.1.2, p.5).” | closeCL confidence: high | Meta Advanced AI Scaling Framework v2 |
| OpenAIShadow mapping OpenAI Preparedness Framework v2 | “"We evaluate whether frontier capabilities create a risk of severe harm through a holistic risk assessment process"; capabilities that pass are tracked as a "Tracked Category or a Research Category" (§2.1).” | closeCL confidence: high | OpenAI Preparedness Framework v2 |
| Frontier Model ForumShadow mapping FMF Risk Taxonomy and Thresholds | “"Most frontier AI frameworks describe structured exercises that developers conduct to identify extreme risks proactively." Threat modelling, "adapted from cybersecurity and national security", is the main method (s2, s2.1, p.7).” | closeCL confidence: high | FMF Risk Taxonomy and Thresholds |
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 chapter splits identification (Commitment 2) from analysis (Commitment 3, five elements including post-market monitoring), and xAI's framework copies that split. Google DeepMind runs identification, inherent risk assessment, mitigation, residual risk assessment and acceptance as one risk management process. Meta organises the work around catastrophic outcomes found in expert workshops. OpenAI's holistic risk assessment ends in a sorting decision (Tracked or Research Category). The EU chapter also says it uses "risk modelling" because "the term 'threat modelling' has a specific meaning in cybersecurity".
