Examples
Examples from published documents
- OpenAI moved persuasion out of its Preparedness Framework and named where it is handled instead: "our Model Spec and policy prohibitions on the use of our tools for political campaigning or lobbying, and our ongoing investigations of misuse of our products" (OpenAI Preparedness Framework v2, Changes in version 2)
- Anthropic's RSP v2.2 "excludes both sophisticated insiders and state-compromised insiders from the ASL-3 Security Standard" (Anthropic Responsible Scaling Policy v3.4, changelog)
Counter-examples
Not to be confused with
- A redaction is not an out-of-scope declaration: redacted material is covered by the assessment but withheld from the public version (see the NIKOLAI element redaction).
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.
- Research Category
- A Research Category, in OpenAI's Preparedness Framework v2, is a capability that does not meet the criteria for a Tracked Category but has "the potential to cause or contribute to severe harm", so OpenAI studies it without attaching threshold-triggered safeguards.
- Source: OpenAI Preparedness Framework v2
- Emerging risk area
- An emerging risk area, in Meta's framework, is an outcome that "may potentially present catastrophic risk and warrant proactive investigation", but for which evaluation methods are "currently too nascent" to bring it into the framework, such as radiological and nuclear risk.
- Source: Meta Advanced AI Scaling Framework v2
Source of record
Where this definition comes from
Meta Advanced AI Scaling Framework v2, §3.2, p.14
“"Some catastrophic harms may not satisfy all four criteria; for example, they may unfold gradually or be partially remediable. These remain serious and are addressed through other safety and integrity processes outside of the scope of this Framework."”
https://ai.meta.com/static-resource/Meta_Advanced-AI-Scaling-Framework-v2OpenAI Preparedness Framework v2, Changes in version 2, item 4
“"Going forward we will handle risks related to persuasion outside the Preparedness Framework, including via our Model Spec and policy prohibitions on the use of our tools for political campaigning or lobbying, and our ongoing investigations of misuse of our products (including detecting and disrupting influence operations)."”
https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf
Crosswalk
How named organisations use this concept
| Organisation | Their term, as published | Match & verification | Source |
|---|---|---|---|
| MetaShadow mapping Meta Advanced AI Scaling Framework v2 | “"These remain serious and are addressed through other safety and integrity processes outside of the scope of this Framework." (§3.2, p.14); emerging areas "may potentially present catastrophic risk and warrant proactive investigation, but existing evaluation approaches are currently too nascent to incorporate these outcomes into our Framework" (§4.2.4).” | exactEQ confidence: high | Meta Advanced AI Scaling Framework v2 |
| OpenAIShadow mapping OpenAI Preparedness Framework v2 | “"Going forward we will handle risks related to persuasion outside the Preparedness Framework, including via our Model Spec and policy prohibitions ..."; "We are moving Nuclear and Radiological capabilities into Research Categories." (Changes in version 2)” | exactEQ confidence: high | OpenAI Preparedness Framework v2 |
| AnthropicShadow mapping Anthropic Risk Report, August 2026 | “"Where laws such as California's Transparency in Frontier Artificial Intelligence Act (aka SB 53) define this or similar terms with specific thresholds, we address those requirements in separate compliance frameworks." (§1, fn 1); section heading "2.17 Considerations outside the scope of this section".” | exactEQ confidence: high | Anthropic Risk Report, August 2026 |
| AnthropicShadow mapping Anthropic Responsible Scaling Policy v3.4 | “Changelog, RSP v2.2 (14 May 2025): "This update excludes both sophisticated insiders and state-compromised insiders from the ASL-3 Security Standard."” Excludes an actor class from a security standard rather than a harm from the framework. | closeCL confidence: high | Anthropic Responsible Scaling Policy v3.4 |
| Google DeepMindShadow mapping Google DeepMind Frontier Safety Framework v3.1 | “"The Framework complements Google's suite of AI responsibility and safety practices, which address other risks in addition to the severe risks in scope of the Framework in accordance with our AI Principles." (s.1.1)” Points to other practices in general terms; no list of excluded risks. | closeCL confidence: high | Google DeepMind Frontier Safety Framework v3.1 |
| Frontier Model ForumShadow mapping FMF Risk Taxonomy and Thresholds | “"These frameworks typically focus on a narrow subset of risks that are severe and irreversible" (s1.1, p.3).” Explains why frameworks exclude most risks; does not itself declare exclusions. | broadBR confidence: medium | 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
Gap
The research found no common format for out-of-scope declarations. Pointers to where excluded risks are handled are usually general ("other safety and integrity processes", "Google's suite of AI responsibility and safety practices"). OpenAI's persuasion exclusion is the most specific pointer found, naming the Model Spec, usage-policy prohibitions and misuse investigations.
