Aleksander Madry says he is returning to OpenAI because safety and security for increasingly capable AI must now be a central priority. In an October 19 announcement, he described what he had thought was a departure as more of a sabbatical. His earlier focus, he said, had been making AI more useful to people.
Madry plans to work with OpenAI chief scientist Jakub Pachocki on those challenges. In his own post, Pachocki confirmed asking Madry to return. He said his focus was the pace and implications of AI progress and how they should inform OpenAI’s strategy and the wider field.
Sam Altman welcomed Madry’s return in his own commented post on October 20. The OpenAI chief described the safety and security challenges around superintelligence as immense. That is an attributed risk assessment, not confirmation that such a system already exists.
The posts establish an announced return and an identified area of work. They do not specify a new title, decision-making powers, additional testing procedures or demonstrated safety results. The personnel decision therefore does not yet establish improved control over AI.