Unverified50% confidenceEventExact time
Anthropic和OpenAI相继发布关于对抗性蒸馏的安全警告，指出攻击者可通过精心构造的提示词从闭源模型中提取核心能力
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/16/2026
First Seen
Valid until: 10/14/2026
Sources
Related Claims
VerifiedOnce open-source model weights are made public, attackers can remove a model's safety restrictions through fine-tuning.71% similarUnverifiedOpenAI对代号为Astra的新模型采取了安全封锁措施，这是该公司首次因模型触发关键网络能力风险阈值而启动最高级别安全响应71% similarUnverifiedAnthropic's cybersecurity safeguards target scenarios involving the prevention of models being misused to generate malicious code, plan attack vectors, or assist in launching cyberattacks against critical infrastructure.70% similarUnverified该事件是OpenAI运行前沿模型网络安全评估时的意外副作用69% similarUnverifiedThe Ministry of Industry and Information Technology issued a security warning regarding a type of AI Agent tool, listing multiple vulnerabilities.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/527991API
curl https://kongchang.com/api/v1/knowledge/claims/527991MCP
get_claim(id=527991)