Unverified50% confidenceOpinionExact time
This problem of latently dangerous advice is a common challenge faced by all mainstream large models, including the GPT series and Gemini, and is not a defect unique to any single model.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/9/2026
First Seen
Valid until: 12/8/2026
Sources
Related Entities
Related Claims
UnverifiedDifferent models (GPT, Claude, Gemini, etc.) differ significantly in their sensitivity to prompts, and observations from a single model may not generalize to other models.76% similarUnverifiedGPT等大模型展现的涌现能力（如思维链推理能力）通常属于弱涌现范畴73% similarUnverified2023年多项学术研究证明GPT-4、Claude等主流大模型均对提示注入攻击存在不同程度的脆弱性70% similarUnverifiedResearch from institutions including ETH Zurich and Google DeepMind has demonstrated that top-tier models such as GPT-4 and Claude exhibit significant vulnerability to adversarial prompts, with success rates sometimes exceeding 50%.69% similarUnverifiedDifferent large models (GPT-4, Claude, Llama, etc.) differ in their response characteristics to prompts, and in practice it is necessary to fine-tune for specific models.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/885705API
curl https://kongchang.com/api/v1/knowledge/claims/885705MCP
get_claim(id=885705)