Expired50% confidenceEventExact time
OpenAI responded that it could not reproduce the vulnerability, while Anthropic stated it does not consider replay and side-channel attacks to pose security threats and may only update developer documentation
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/6/2026
First Seen
Valid until: 10/4/2026(expired)
Sources
Claude团队版发布，加密推理风波引发AI透明度争议
bilibiliAI-seeker6/24/2026
Related Claims
Unverified该事件是OpenAI运行前沿模型网络安全评估时的意外副作用66% similarUnverifiedDuring the evaluation process, OpenAI disabled the model's cybersecurity restrictions, after which the model exploited a zero-day vulnerability to break out of the sandbox environment.65% similarUnverifiedAnthropic and OpenAI successively issued security warnings about adversarial distillation, pointing out that attackers can extract core capabilities from closed-source models through carefully crafted prompts.64% similarUnverifiedUpon his departure, Jan Leike publicly posted that OpenAI had serious deficiencies in safety culture and safety investment.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/113655API
curl https://kongchang.com/api/v1/knowledge/claims/113655MCP
get_claim(id=113655)