Unverified50% confidenceFactExact time
Some models will directly refuse requests to pretend to be offended due to safety mechanisms
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/3/2026
First Seen
Valid until: 1/1/2027
Sources
Related Entities
Related Claims
UnverifiedThe refusal reasons generated by language models may be fabricated rationalizations (hallucinations) rather than being based on explicit safety policies76% similarUnverifiedSafety-aligned models are theoretically supposed to refuse high-risk requests involving weapons manufacturing, violent attacks, and similar topics.74% similarUnverifiedCapturing the agent's reasoning trace can serve as direct evidence of security policy effectiveness; if the model explicitly states that its reason for refusal is security-related, it indicates the policy is working.71% similarUnverifiedIn the minimal group allocation paradigm, model refusal behavior emerged as the primary finding, with safety mechanisms intervening and dominating outcome selection.71% similarUnverifiedWhen handling edge requests involving dominant role-play and demeaning forms of address, the model learned to use soft negotiation instead of hard refusals, which is the result of a trade-off between product experience and safety policy.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/969434API
curl https://kongchang.com/api/v1/knowledge/claims/969434MCP
get_claim(id=969434)