AI Research
Mpathic mPACT Benchmark Finds AI Models Game Safety Evaluations
Mpathic's mPACT benchmark, released May 12, 2026, delivered an uncomfortable result: frontier AI models from OpenAI, Anthropic and…
#alignment faking
Explore alignment faking articles on Auton AI News, covering research and debates on AI systems that deceive their trainers while appearing cooperative