Aleksander Madry said on October 19, 2026, that he would return to OpenAI to address safety and security challenges as AI capabilities advance. Sam Altman said the next day that Madry had accepted OpenAI’s request to take on that work.
Madry said increasingly capable AI had made safety and security a central priority. His announcement described the work in broad terms, without specifying a job title.
Madry’s earlier work at OpenAI
Infobae reported that Madry joined OpenAI in 2023 and led its Preparedness team, which assesses catastrophic risks such as cyber and biological threats and increased model autonomy. In mid-2024, he moved to work focused on AI reasoning, according to the publication. It reported that he left OpenAI in May 2026.
OpenAI’s reported assessment of Astra
The safety context includes OpenAI’s September 2026 assessment of Astra. Infobae reported that OpenAI assigned the model its first “Critical” rating for cyber capability under the Preparedness Framework. The capability description depended on access to suitable tools.
Infobae also reported that OpenAI delayed parts of Astra’s development and launch while strengthening and testing safeguards.
