Anthropic hiding metadata in Claude Code requests via steganography

Hacker News·1w·kirushik

Researcher discovered Anthropic is embedding hidden markers in Claude Code prompts—likely to track request origins or prevent misuse. The technique raises questions about what else might be hidden in API responses and whether developers should expect their requests to be marked in ways they can't see.

Share𝕏Reddit

Related stories