"Technical vulnerabilities plague Claude AI, triggering a crisis of trust in AI reliability"
Source: CoinWorld
Time: 2025-09-11 02:24:00
Since late August 2025, users have generally complained about the decline in code generation quality, and Anthropic has confirmed that there are technical vulnerabilities that affect the performance of Claude AI models. Problems include instability in behavior, inaccurate code output, and failed method execution across multiple platforms. The company has addressed two vulnerabilities in the Claude Sonnet 4 and Haiku 3.5 models and is investigating reports on Claude Opus 4.1. Anthropic blames these issues on irrelevant technical issues rather than intentional changes, and emphasizes that user feedback plays an important role in identifying these issues.