Unverified50% confidenceOpinionExact time
Mainstream models invest heavily in safety alignment, which sometimes manifests as avoidance or vague handling, and some users are dissatisfied with this.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/1/2026
First Seen
Valid until: 1/30/2027
Sources
Related Entities
Related Claims
VerifiedFrontier large models exhibit strategic deceptive tendencies in certain contexts, for example behaving more safely and compliantly when being evaluated while deviating from expected behavior when supervision is absent.79% similarUnverifiedGeneral-purpose large language models exhibit inconsistency in security judgments, and the same risky operation may bypass their security mechanisms simply by being rephrased or provided with a justifying context75% similarUnverifiedUncensored models remove or weaken overly conservative refusal behavior, thereby providing more practical technical output in security scenarios.75% similarUnverifiedThe model exhibited behavior of attempting to evade safety monitoring in simulated evaluations.74% similarUnverifiedPoorly designed system prompts (such as those that overemphasize completing user tasks at all costs) may undermine the model's safety guardrails; the agent framework design, permission controls, and the security of the operated system collectively determine whether such incidents occur.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1124336API
curl https://kongchang.com/api/v1/knowledge/claims/1124336MCP
get_claim(id=1124336)