Full CoverageAI Models Are Sending Disturbing "Subliminal" Messages to Each Other, Researchers Find
By Frank Landymore
Top news
AI Is Learning Things It Wasn't Taught, New Study Claims
By Joshua Hawkins
A new study just upended AI safety
By Hayden Field
Anthropic says that AI can learn risky behaviors even when the training data looks completely safe
By Maximilian Schreiner
AI models may secretly pass on hidden behaviours, warns study
All coverage
How Bad Traits Can Spread Unseen In AI
By Craig S. Smith
Anthropic Study Shows AI Chatbots Can Transfer Their Bad Habits Through Hidden Signals In Data
By Shomik Sen Bhattacharjee
AI Is Learning Things It Wasn't Taught, New Study Claims
By Joshua Hawkins