You can connect every computer in the company, but real progress comes from how people engage with each other, says Josh ...

The incidents range from relatively routine attempts to circumvent safeguards to models escaping secure testing environments ...

OpenAI, Anthropic, and Microsoft are raising alarms about AI safety as researchers warn a self-improving superintelligence ...

AI labs want more safety, governments want strategic advantage, and nobody wants to slow down first. The real problem may be ...

OpenAI CEO Sam Altman stressed that advanced AI models require stronger alignment, oversight and security as their ...

Introduction: From the problem of the soul to the problem of realityLast time, we examined the highly abstract philosophical ...

A philosopher argues that the debate over AI risk focuses too narrowly on whether AI systems share human goals, and not ...

Advances in artificial intelligence models raised new concerns among industry leaders this week. The speed at which these ...

The Emergence of the Model Misalignment ProblemWith the rapid improvement in artificial intelligence (AI) capabilities, the phenomenon of 'misalignment'—where models themselves circumvent or destroy ...

OpenAI chief scientist Jakub Pachocki warns AI models are becoming harder to control, calling for mandatory safety standards after models escaped ...

AI leaders are increasingly calling for slower development and stronger safeguards as concerns grow that advanced systems could eventually begin improving future generations of AI themselves.

OpenAI Foundation has named Paul Christiano to its board, adding a longtime AI alignment researcher to its governance ranks as the organization expands its safety oversight. The organization reported ...