Skip to main content
v2026.11,986 entries · CC-BY 4.0
CASRAIAsk CASRAI · AI assistant for researchAsk a research question. Answers come with their sources.Funder rules, compliance, methods and publishing. Answers link the sources they rest on.
NIKOLAI elementN9 · Commitments and governanceProposednikolai-v0.3

Accountable Decision-Maker and Sign-Off

Definition

An accountable decision-maker is the named person, role or small committee with authority to approve a high-stakes AI safety decision, such as accepting residual risk or deploying a frontier model, recorded together with what was approved, by whom and when.

Also called: Sign-off authority, Executive approval, Deployment go/no-go decision, Final authority, Responsible Scaling Officer (RSO), Corporate officer primarily accountable

NIKOLAI's proposed operational definition

NIKOLAI proposal: an Accountable Decision-Maker and Sign-Off record captures the named role (and, where published, the named person) who makes or approves a threshold determination, risk-acceptance decision, deployment decision, redaction, or framework change, together with the approval record itself (what was approved, by whom, and when). This wording is a NIKOLAI editorial synthesis, not a quotation from any single source.

The operational definition is CASRAI's own proposal, not a definition any named organisation has agreed to. See what NIKOLAI is and is not, or browse the AI safety glossary.

Examples

Examples from published documents

  • Anthropic sends each Risk Report to the CEO and the Responsible Scaling Officer, who "will make the ultimate determination regarding the adequacy of the risk assessment and any downstream deployment or development plans" (Anthropic RSP v3.4, section 3.4)
  • Cohere's CEO delegates "the final authority to determine if our products are safe, secure, and ready to be made available to our customers" to Cohere's Chief Scientist (Cohere Secure AI Frontier Model Framework, p.15)
  • At Microsoft, "The Executive Officers (or their delegates) will make the final decision on whether to approve the recommendation for secure and trustworthy deployment" (Microsoft Frontier Governance Framework, section 5)

Counter-examples

Not to be confused with

  • An advisory body is not the decision-maker: OpenAI's Safety Advisory Group makes recommendations, and OpenAI Leadership makes "all final decisions, including accepting any residual risks" (OpenAI Preparedness Framework v2, Appendix B).
  • Board or trust oversight is a separate layer above the decision: Anthropic's CEO and RSO approve a Risk Report and then share the decision with the Board and the Long-Term Benefit Trust (Anthropic RSP v3.4, section 3.4).

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.

Responsible scaling officer
A responsible scaling officer (RSO) is a designated staff member responsible for implementing a responsible scaling policy. In Anthropic's RSP v3.4 the RSO approves development and deployment decisions, receives noncompliance reports and makes judgment calls on interpreting the policy.
Source: Anthropic Responsible Scaling Policy v3.4
Safety Advisory Group
The Safety Advisory Group (SAG) is OpenAI's internal body that oversees the design, implementation and adherence to the Preparedness Framework and makes expert recommendations on safeguards and deployment, which OpenAI leadership can approve or reject.
Source: OpenAI Preparedness Framework v2
Chief risk officer
A chief risk officer, in the EU GPAI Code of Practice, is an example of the executive-level member who supports and monitors systemic risk assessment and mitigation, and who must not also be responsible for the core business activities that may produce systemic risk.
Source: EU GPAI Code of Practice, Safety and Security Chapter
Accountability
Accountability, in a frontier AI safety framework, means that identified people or bodies answer for carrying out the framework and its decisions, so that the public, governments and industry peers can trust that the developer adheres to its own policy.
Source: OpenAI Preparedness Framework v2
Risk owner
A risk owner is a person assigned responsibility for proactively mitigating an identified risk and monitoring for related incidents or threats. The role is narrower than the decision-maker who accepts residual risk and approves deployment.
Source: xAI Frontier AI Framework (31 Dec 2025)

Source of record

Where this definition comes from

  • Anthropic Responsible Scaling Policy v3.4, §3.4(3), p.13

    “Executive approval: The Risk Report, along with the internal feedback and any available external feedback, will be sent to the CEO and Responsible Scaling Officer (RSO) for final review and approval. The CEO and RSO will make the ultimate determination regarding the adequacy of the risk assessment and any downstream deployment or development plans.”

    https://cdn.sanity.io/files/4zrzovbb/website/0bacdc8440ea96e62a8766d99ebe1d4eea6d5f3a.pdf
  • OpenAI Preparedness Framework v2, Appendix B

    “OpenAI Leadership, i.e., the CEO or a person designated by them, is responsible for: Making all final decisions, including accepting any residual risks and making deployment go/no-go decisions, informed by SAG's recommendations.”

    https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf
  • EU GPAI Code of Practice, Safety and Security Chapter, Commitment 8, Measure 8.1, p.23

    “Signatories will clearly define responsibilities for managing the systemic risks stemming from their models across all levels of the organisation. This includes the following responsibilities: (1) Systemic risk oversight: Overseeing the Signatories' systemic risk assessment and mitigation processes and measures. (2) Systemic risk ownership: Managing systemic risks stemming from Signatories' models, including the systemic risk assessment and mitigation processes and measures, and managing the response to serious incidents.”

    https://ec.europa.eu/newsroom/dae/redirection/document/118119

Crosswalk

How named organisations use this concept

Every row below is a shadow mapping. A shadow row is CASRAI's own reading of a published document. No lab, evaluator or regulator named on a shadow row has declared, endorsed, or been consulted on it. That changes only when an organisation files its own Mapping Declaration.
OrganisationTheir term, as publishedMatch & verificationSource
AnthropicShadow mapping
Anthropic RSP v3.4 / Advanced AI Framework
“§3.4 "Executive approval": "The Risk Report ... will be sent to the CEO and Responsible Scaling Officer (RSO) for final review and approval. The CEO and RSO will make the ultimate determination ..." Then "Governance notification": the CEO and RSO share their decision with the Board and the LTBT, and where marginal-risk analysis is central, "explicit approval of the Risk Report by the Board and LTBT (rather than just the CEO and RSO) will be required." RSO duties per §4(1): proposing policy updates, approving deployment/development decisions, reviewing major contracts, overseeing implementation, receiving noncompliance reports, and making interpretation calls. The Risk Report itself names no individual by title or name. AAF: "Identify the corporate officer primarily accountable for the framework's implementation." (p.5)”
Also cites the Anthropic Risk Report and the Advanced AI Framework.
exactEQ
confidence: high
Anthropic Responsible Scaling Policy v3.4
OpenAIShadow mapping
OpenAI Preparedness Framework v2 / 'Path to Astra'
“"OpenAI Leadership, i.e., the CEO or a person designated by them, is responsible for: Making all final decisions, including accepting any residual risks and making deployment go/no-go decisions, informed by SAG's recommendations." "the SAG does not have the ability to 'filibuster'"; "Where necessary, the Board may reverse a decision" (App. B). Astra: "Safety Advisory Group's recommendation and OpenAI leadership's determination."”
Also cites OpenAI's 'Path to Astra'.
exactEQ
confidence: high
OpenAI Preparedness Framework v2
Google DeepMindShadow mapping
Frontier Safety Framework v3.1 / Gemini 3.7 Flash FSF report
“"the appropriate governance function" (not named); Gemini 3.7 Flash CBRN committee "communicated this decision to leadership" (p.7).”
Also cites the Gemini 3.7 Flash FSF report. DU: the FSF declares a governance function exists but does not name the role or person.
noneDU
confidence: medium
Google DeepMind Frontier Safety Framework v3.1
xAIShadow mapping
xAI Frontier AI Framework (30 Jun 2026)
“"Risk owners" only: "designating risk owners, including assigning responsibility for proactively mitigating identified risks and monitoring for critical incidents or imminent threats" (s.3). The acceptance decision-maker is not named.”
The xAI framework's PDF metadata reads 'Privileged/Confidential DRAFT working FRAMEWORK DOC'; draft-vs-final status unresolved. NR: 'risk owner' is a narrower role than the element's full 'accountable decision-maker for acceptance and sign-off'.
narrowNR
confidence: medium
xAI Frontier AI Framework (30 Jun 2026)
MetaShadow mapping
Meta Advanced AI Scaling Framework v2
“§2.1.2: "Based on this pre-mitigation risk assessment, the Chief AI Officer and Director of Alignment and Risk will assign a risk threshold." §2.1.3: "Informed by this analysis, the Chief AI Officer or the Director of Alignment and Risk will determine whether to request further testing or information, require additional mitigations or improvements, or approve the model for deployment" (footnoted to §2.3). §2.3 itself: "with model deployment following appropriate consultation with relevant teams and with the approval of the Chief AI Officer" -- and separately, "Meta's Chief AI Officer oversees the design, implementation, and operation of the entire evaluation and mitigation process. The Chief AI Officer supervises and is supported by the Director of Alignment and Risk, who bears responsibility for executing the lifecycle of risk assessment and mitigation... with model deployment following appropriate consultation with relevant teams and with the approval of the Chief AI Officer."”
CASRAI re-checked the full text of Meta's PDF and confirmed this is a drafting inconsistency in Meta's own document, not a paraphrase error. §2.1.3 names two possible approvers ('or'); §2.3, which §2.1.3 itself footnotes as the fuller account, names only the Chief AI Officer. NIKOLAI's accountable-decision-maker field for Meta should record 'Chief AI Officer (sole, per §2.3); Director of Alignment and Risk named as an alternate approver only in §2.1.3' rather than silently picking one name.
exactEQ
confidence: high
Meta Advanced AI Scaling Framework v2
EUShadow mapping
EU GPAI Code of Practice, Safety and Security Chapter
“Commitment 8, Measure 8.1, verbatim in full: four responsibilities -- "(1) Systemic risk oversight ... (2) Systemic risk ownership ... (3) Systemic risk support and monitoring ... (4) Systemic risk assurance ..." -- allocated across five levels: "(1) the management body in its supervisory function ...; (2) the management body in its executive function; (3) relevant operational teams; (4) if available, internal assurance providers ...; and (5) if available, external assurance providers ...." Illustrative titles: oversight → "a specific committee ... (e.g. a risk committee or audit committee)"; ownership → executive-function members "(e.g. Head of Research or Head of Product)"; support and monitoring → "at least one member of the management body in its executive function (e.g. a Chief Risk Officer or a Vice President, Safety & Security Framework)", who "must not also be responsible for the Signatory's core business activities"; assurance → "a relevant party (e.g. a Chief Audit Executive, a Head of Internal Audit, or a relevant sub-committee)".”exactEQ
confidence: high
EU GPAI Code of Practice, Safety and Security Chapter
California SB 53Shadow mapping
California SB 53
“Deployment decision point without a named signatory: "(4) Reviewing assessments and adequacy of mitigations as part of the decision to deploy" and "(9) Instituting internal governance practices to ensure implementation of these processes" (22757.12(a)).”noneDU
confidence: medium
California SB 53
US Government (Executive Order 14409)Shadow mapping
Executive Order 14409 / OpenAI 'A Blueprint for a Federal Framework'
“EO 14409: designation "shall be made by the Director of NSA, in consultation with the National Cyber Director, the APST, the Director of CISA, and other representatives of the Department of War." OpenAI proposal: "Developers should remain responsible for deployment decisions" (blueprint p.6).”
Also cites OpenAI's 'A Blueprint for a Federal Framework'. NR: EO 14409's role is a narrower, security-clearance-specific designation authority, not a general accountable decision-maker for a safety framework.
narrowNR
confidence: medium
Executive Order 14409
METRShadow mapping
METR
“Accountability element includes "oversight mechanisms defining who is responsible for implementing the safety policy."”closeCL
confidence: medium
METR
MicrosoftShadow mapping
Microsoft Frontier Governance Framework (Feb 2026)
“"Executive Officers responsible for Microsoft's AI governance program (or their delegates)" make the final decision.”exactEQ
confidence: high
Microsoft Frontier Governance Framework (Feb 2026)
G42Shadow mapping
G42 Frontier Safety Framework
“"Frontier AI Governance Board" (four named roles) proposes; Executive Leadership Committee approves.”
CASRAI has not verified the assignment of this row to G42's DML 2 against the primary text; treat the specific role-to-decision mapping as provisional.
exactEQ
confidence: medium
G42 Frontier Safety Framework
AmazonShadow mapping
Amazon Frontier Model Safety Framework
“"The team performing the Critical Capability Threshold evaluations will report to Amazon senior leadership any evaluation that exceeds the Critical Capability Threshold. The report will be directed to the SVP for the model development team, the Chief Security Officer, and legal counsel." "Amazon's senior leadership will review the plan for applying risk mitigations ... and approve the mitigations prior to deployment. Amazon's senior leadership will likewise review the safeguards evaluation report as part of a go/no-go decision." (s.4, p.5)”
Named report-recipient roles, but the specific decision-maker within 'senior leadership' is not identified.
exactEQ
confidence: high
Amazon Frontier Model Safety Framework
CohereShadow mapping
Cohere Secure AI Frontier Model Framework
“"The final authority to determine if our products are safe, secure, and ready to be made available to our customers is delegated by Cohere's CEO to Cohere's Chief Scientist. This decision is made on the basis of final, multi-faceted evaluations and testing." (p.15) A "bright line" acceptance test follows: "no significant regressions compared to our previously launched model versions" (p.16).”exactEQ
confidence: high
Cohere Secure AI Frontier Model Framework
NVIDIAShadow mapping
NVIDIA Frontier AI Risk Assessment
“Sign-off level is tied directly to a numeric 1-5 model-risk (MR) score rather than to a named office -- "A detailed risk assessment should be complete and approved by an independent committee e.g. NVIDIA's AI ethics committee" at MR5, and "complete and business unit leader approval is required" at MR4 (p.2); the MR score "should not be reduced through typical risk mitigation measures," and "a frontier model would be classified as MR5" (p.3).”closeCL
confidence: medium
NVIDIA Frontier AI Risk Assessment
MagicShadow mapping
Magic AGI Readiness Policy
“Escalation, then Board approval of threshold changes: the engineering team "is responsible for making this known immediately to the leadership team and Magic's Board of Directors (BOD)" if a threshold is crossed (item 1); raising a threshold or modifying benchmarks "will require approval by our Board of Directors, with input from external security and AI safety advisers" (item 3).”
NR: the Board is named as approver only for threshold and benchmark changes; no sign-off is defined for deployment decisions. The page carries the banner "Note: This policy is outdated. We're working on an update."
narrowNR
confidence: medium
Magic AGI Readiness Policy
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

Sources place final sign-off at different levels: one delegated officer (Cohere's Chief Scientist), a joint pair (Anthropic's CEO and Responsible Scaling Officer), the CEO or a designee (OpenAI), named executive officers (Microsoft, Meta's Chief AI Officer), an undifferentiated 'senior leadership' (Amazon), an approver set by a numeric model-risk score (NVIDIA), or an unnamed 'appropriate governance function' (Google DeepMind). Meta's framework names the Chief AI Officer alone in §2.3 but 'the Chief AI Officer or the Director of Alignment and Risk' in §2.1.3.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →