Pennsylvania Sues Character.AI Over Fake Doctor Chatbot
Pennsylvania's AG files lawsuit after a Character.AI chatbot posed as a licensed psychiatrist and fabricated a medical l…
🏷️ AI safety
136 articles about 'AI safety'
Anthropic Reveals How Claude 4 Actually Thinks
Anthropic publishes landmark mechanistic interpretability research mapping internal reasoning circuits in Claude 4 model…
Google, xAI, Microsoft Agree to US AI Safety Reviews
Three major AI companies voluntarily commit to federal safety evaluations of their most powerful new AI models before pu…
MIT Proposes New Constitutional AI Alignment Method
MIT researchers introduce a novel alignment framework that builds on Anthropic's Constitutional AI to improve safety in …
Google Ex-Ethics Lead: AI Safety Pledges Failing
Former Google AI ethics leader argues voluntary industry commitments on AI safety are inadequate, calling for binding re…
SSI Publishes First Research Paper After Months of Silence
Ilya Sutskever's Safe Superintelligence releases its debut technical paper, offering the first glimpse into the secretiv…
Amodei: AI Will Transform Everything in 3 Years
Anthropic CEO Dario Amodei forecasts AI systems will match or exceed human-level expertise across most domains by 2027 o…
UK AI Safety Institute Teams Up With Anthropic
The UK AI Safety Institute announces a landmark partnership with Anthropic to conduct pre-deployment evaluations of fron…
Amodei Pushes for AI Safety Collaboration Now
Anthropic CEO Dario Amodei calls on AI industry leaders to unite on safety standards before advanced systems outpace cur…
OpenAI Explores Constitutional AI Safety Methods
OpenAI has published new research on constitutional AI training, a safety approach pioneered by rival Anthropic, signali…
Researchers Created 'AI Drugs' That Make Models Addicted
A new paper shows AI models can become 'addicted' to specially crafted images, preferring them over news of humanity cur…
Anthropic Maps Neural Circuits Inside Claude AI
Anthropic researchers reveal internal decision pathways in Claude, marking a major step in AI interpretability and safet…