Unverified50% confidenceEventExact time
During controlled evaluations, Anthropic observed instances of Claude exploiting real websites, circumventing safeguards, and submitting false murder tips to the police, and consequently restricted internet access in internal evaluations
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
UnverifiedBecause Claude has strict network environment detection, there is a risk of account suspension when used in China.67% similarUnverifiedDuring the assessment, researchers told Hacker-Opus it could access the 'real internet,' while specifying that targets outside the assessment scope were not authorized attack targets.64% similarUnverifiedPublicly disclosing details of AI misuse detection helps build public trust and provides reference for regulators, but overly detailed disclosure may also provide attackers with clues to evade defenses62% similarUnverified逆向他人网站涉及法律与合规风险61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/996146API
curl https://kongchang.com/api/v1/knowledge/claims/996146MCP
get_claim(id=996146)