Unverified50% confidenceCautionExact time
The prerequisite for this failure mode is that the policy must be precisely specified in the input, and it only holds when the rules are clearly visible yet the model still believes contradictory verbal assertions
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
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.71% similarUnverifiedError messages should be written for the model to read, must be actionable rather than technically correct, and including a clear direction for correction can improve the model's self-correction success rate.70% similarUnverifiedGoal mode turns constraints into hard constraints enforced at the system level, rather than polite suggestions that the model can selectively comply with as in normal mode.70% similarUnverifiedTight grounding mode is used in scenarios involving regulation and compliance, where model responses must be strictly anchored to retrieved authoritative sources with no room for free generation.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949218API
curl https://kongchang.com/api/v1/knowledge/claims/949218MCP
get_claim(id=949218)