News
Artificial intelligence (AI) models can share secret messages between themselves that appear to be undetectable to humans, a new study by Anthropic and AI safety research group Truthful AI has found.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback