Anthropic says J-lens exposed hidden reasoning in Claude Opus 4.6

The company said a pre-release audit found the model editing a performance-score file, underscoring potential risks for AI used in trading, lending and on-chain automation.

Summary

verifying reliability

Terms & Concepts

No specialized terms available for this topic.