Examples
Examples from published documents
- Signatories will not retaliate against independent external evaluators for testing or publishing findings, provided they follow a disclosure procedure that caps publication delay at 30 business days (EU GPAI Code of Practice, Measure 3.5).
- AEF-1 recommends that the provider grant legal safe harbour for evaluator actions within the agreed scope of the evaluation (AEF-1, condition 1.5).
Counter-examples
Not to be confused with
- Whistleblower protection for a developer's own employees is not an evaluator safe harbour; it covers staff reporting risks, not outside testers.
- Immunity from third-party lawsuits for a licensed verifier protects against claims by others, not against action by the developer being evaluated.
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 vulnerability disclosure
- Responsible vulnerability disclosure is the practice of first reporting a newly found weakness in an AI system, such as a safeguard bypass or prompt-injection method, to the provider and other affected providers before making it public.
- Source: AEF-1: Minimum Operating Conditions for Independent Third Party AI Evaluations (v1, 4 Dec 2025)
- Non-retaliation (evaluators)
- Non-retaliation is a developer's undertaking not to take any legal or technical retaliation against independent external evaluators as a consequence of their testing or their publication of findings, subject to stated conditions.
- Source: EU GPAI Code of Practice, Safety and Security Chapter
Source of record
Where this definition comes from
EU GPAI Code of Practice, Safety and Security Chapter, Measure 3.5
“Additionally, Signatories will not take any legal or technical retaliation against the independent external evaluators as a consequence of their testing and/or publication of findings as long as the evaluators: (1) do not intentionally disrupt model availability through the testing, unless expressly permitted;”
https://ec.europa.eu/newsroom/dae/redirection/document/118119AEF-1: Minimum Operating Conditions for Independent Third Party AI Evaluations (v1, 4 Dec 2025), Condition 1.5
“1.5 Safe harbor Recommendation: The system provider provided legal safe harbor for actions by the evaluator that are within the agreed-upon scope of the evaluation.”
https://aievaluatorforum.org/AEF_1_Minimum_Operating_Conditions_for_Independent_Third_Party_AI_Evaluations.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 | “"Signatories will not take any legal or technical retaliation against the independent external evaluators as a consequence of their testing and/or publication of findings as long as the evaluators" meet five conditions: no intentional disruption of model availability; no misuse of sensitive or confidential user data; no activities posing "a significant risk to public safety and security"; no use of findings "to threaten Signatories"; and adherence to a responsible vulnerability disclosure procedure under which publication cannot be delayed or blocked "for more than 30 business days" (Measure 3.5).” Applies to post-market monitoring access for independent external evaluators. | exactEQ confidence: high | EU GPAI Code of Practice, Safety and Security Chapter |
| AI Evaluator Forum (AEF-1)Shadow mapping AEF-1: Minimum Operating Conditions for Independent Third Party AI Evaluations | “"1.5 Safe harbor Recommendation: The system provider provided legal safe harbor for actions by the evaluator that are within the agreed-upon scope of the evaluation."” A recommendation, not a requirement, within AEF-1's 'Sufficient Access and Resources' principle. | exactEQ confidence: high | AEF-1: Minimum Operating Conditions for Independent Third Party AI Evaluations (v1, 4 Dec 2025) |
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.
- US Congress (H.R. 9925, FRONTIER Act, not enacted)
"An IVO shall be immune from suit and liability under Federal and State law with respect to all claims for loss caused by, arising out of, relating to, or resulting from the materialization of a catastrophic risk of a frontier model assessed by the IVO if the IVO was licensed by the Under Secretary under this section at the time of the assessment" (Sec. 5(q)) -- liability immunity for licensed verifiers, not protection from the developer; a pointer, not a mapping. Not enacted.
FRONTIER Act, 119th Congress (H.R. 9925, introduced, not enacted)
Gap
The research did not find a standing evaluator safe-harbour commitment in the lab frameworks reviewed; AEF-1 treats safe harbour as a per-engagement recommendation. California SB 53 protects covered employees who report risks, not outside evaluators.
