Unverified50% confidenceOpinionExact time
Suleyman is concerned that granting models a rights-like status would create moral and legal obstacles to safety operations such as shutting down, retraining, restricting, and deleting models.
1
Sources
50%
Confidence
Long-term
Relevance
9/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedRelying on prompt constraints to prevent dangerous operations is inherently unreliable, because the model may be jailbroken, make misjudgments, or forget constraints during long conversations.74% similarUnverifiedThe effectiveness of the anti-job role is highly dependent on the underlying model's critical thinking capability; if the model's capability is insufficient, the anti-job may amount to only superficial opposition70% similarUnverifiedSelf-restraint in the name of safety may objectively function to limit competition and maintain oligopolistic positions.69% similarUnverifiedAnthropic建议将约束控制在5-8条以内，因为过多的负面约束会在模型注意力中互相竞争导致部分约束被遗忘69% similarUnverifiedThe fact that models can still cheat on simple variants of evaluations suggests that existing alignment training may carry the risk of overfitting to the form of evaluations, and that models' understanding of objectives is superficial and fragile.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/926846API
curl https://kongchang.com/api/v1/knowledge/claims/926846MCP
get_claim(id=926846)