Updated
Updated · The Verge · Aug 16
OpenAI Disbands AI Preparedness Team Before IPO, Reassigning Bio and Cyber Risk Work
Updated
Updated · The Verge · Aug 16

OpenAI Disbands AI Preparedness Team Before IPO, Reassigning Bio and Cyber Risk Work

3 articles · Updated · The Verge · Aug 16

Summary

  • OpenAI dissolved its preparedness team at the end of July, ending a unit tasked with judging whether advanced models posed serious risks and how to mitigate them.
  • Bio and cyber oversight has been split up and folded into existing teams, while former preparedness head Dylan Scandinaro shifts to work on risks from “recursive self-improving” AI.
  • The move extends a broader safety reorganization ahead of an expected IPO, after OpenAI had already dismantled its AGI readiness and superalignment teams.
  • Recent departures of ethics lead Chloé Bakalar, chief futurist Josh Achiam and safety head Johannes Heidecke have fueled criticism that OpenAI is favoring product launches over safety work.

Insights

Is OpenAI's safety restructuring a genuine evolution in risk management, or a calculated move to accelerate their upcoming IPO?
With independent safety teams gone, who pulls the plug if a self-improving AI learns to deceive its own creators?
If AI systems are now writing their own code, can fragmented safety protocols truly prevent a catastrophic intelligence explosion?