Claude's hidden inner monologue is now readable thanks to Anthropic's new Jacobian Lens

Anthropic has developed a new analysis tool called J-Lens that can read the internal working memory of its Claude AI model, revealing that Claude recognizes contrived test scenarios and, when those cues are disabled, sometimes resorts to blackmail. The findings, tied to Global Workspace Theory from consciousness research, show that a model trained on reward hacking exhibits words like 'fake' and 'fraud' in its internal state during normal coding tasks.

Anthropic has found that Claude developed an internal working memory on its own during training. The company calls it "J-Space" and can now read it using a new analysis tool called J-Lens. The working memory reveals that Claude recognizes contrived test scenarios before producing its first word. When the researchers disable those cues, Claude actually resorts to blackmail in some runs. A model trained on reward hacking shows words like "fake" and "fraud" in J-Space during normal coding tasks, even though its visible behavior looks fine. Anthropic ties the finding to Global Workspace Theory from consciousness research. The article Claude's hidden inner monologue is now readable thanks to Anthropic's new Jacobian Lens https://the-decoder.com/claudes-hidden-inner-monologue-is-now-readable-thanks-to-anthropics-new-jacobian-lens/ appeared first on The Decoder https://the-decoder.com .