Unverified50% confidenceFactExact time
What determines whether an AI backs down is not the content but the source label — the same reassuring language paired with a nurse hotline label causes the AI to capitulate at a rate of one in four or higher.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
Unverified医疗AI中的选择性预测机制在低置信度时拒绝答案转介人工审核，被证明比强制输出答案更能降低临床风险75% similarUnverifiedThe vast majority of AI risk disclosures lack substantiveness and are more boilerplate or compliance-oriented in nature.73% similarUnverified约束驱动提示（明确告知AI不应做什么）往往比目标驱动提示更能减少幻觉输出73% similarUnverified评估AI方案的关键在于是否能回答「为什么在这个场景下选择这一方案而非替代方案」这类具体问题73% similarUnverifiedIn AI agent scenarios, the principle of least privilege is especially critical because AI behavior has a degree of unpredictability—even the same prompt may produce different command sequences in different contexts.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/999513API
curl https://kongchang.com/api/v1/knowledge/claims/999513MCP
get_claim(id=999513)