Top OpenAI Catastrophic Risk Official Steps Down Abruptly

a year ago

Joaquin Quiñonero Candela, OpenAI's top safety staffer, steps down from his role as Head of Preparedness to become an intern focusing on healthcare applications of AI.
OpenAI restructures its safety organization, consolidating governance under the Safety Advisory Group (SAG) led by Sandhini Agarwal.
The Preparedness team, established to mitigate catastrophic risks from AI, has seen leadership changes, including the reassignment of Aleksander Mądry and now Candela's departure.
OpenAI faces an exodus of safety leadership, including cofounder John Schulman, safety lead Lilian Weng, and Superalignment team leads Ilya Sutskever and Jan Leike.
Miles Brundage, Senior Advisor for AGI readiness, resigns, citing concerns over OpenAI's readiness for AGI and restrictions on public discussions about AI risks.
OpenAI's Safety and Security Committee (SSC) undergoes changes, with board member Zico Kolter joining as chair and Sam Altman no longer on the committee.
Concerns grow over OpenAI's commitment to safety, including reduced AI model safety testing time and the release of GPT-4.1 without a corresponding safety report.
Former employees and researchers criticize OpenAI for reducing safety commitments and lacking transparency in safety leadership and procedures.
Global implications of AI safety are highlighted, with calls for more attention to economic and strategic risks posed by AI developments.

Hasty Briefsbeta