AI · 2h ago
Anthropic Removes Steganography Code Targeting Chinese AI Rivals
Anthropic is disabling a covert steganography system embedded in its AI models to detect Chinese competitors' usage. The code, which hid watermarks in model outputs, was designed to identify unauthorized access by Chinese firms. Anthropic confirmed the removal, citing a shift in strategy.
Meridian48 take
The move suggests Anthropic is prioritizing trust over surveillance, but the initial deployment of secret detection code raises questions about transparency in AI model distribution.
Read the full reporting
Anthropic is removing its covert code for catching Chinese competitors →
The Register
anthropicsteganography