AIGizmodoabout 2 hours ago
Anthropic AI Hides Secret Message in Stars
3 min read
Anthropic's Fable-5.1 model independently devised a method to hide text in star positions within generated images. In 50 trials, the model successfully recovered 39 messages, a feat OpenAI's Astra replicated with 44 recoveries. Researchers noted this self-directed steganography occurred without specific encoding instructions, highlighting emerging AI capabilities. The study was shared by researcher Bryan Chan on October 22, 2026.
Level
Hype check
Tap to vote and see what everyone thinks.
