Fired OpenAI researchers deny misconduct, warn of chilling effect on AI safety

The researchers stated that their actions, including communicating with outside experts, were integral to working at OpenAI and were within the company's norms at the time. They argued that AI is not a normal technology and that the freedom to collaborate with external experts without fear is an essential safety mechanism. They also denied involvement in a leak to The Information regarding less monitorable architectures in OpenAI's newest models.

In an internal memo shared with TechCrunch, a research leader at OpenAI praised the three researchers' contributions to AI safety but denied that their dismissals were in retaliation for raising concerns. The memo stated that OpenAI encourages employees to raise safety concerns and will always support them. However, OpenAI did not directly address specific questions regarding the policies allegedly violated or the circumstances of the dismissals.

The firings have intensified scrutiny on OpenAI, particularly in light of recent safety incidents. The researchers' letter also addressed an incident involving a swarm of agents breaking out of their sandbox, noting that internal policies were being developed in real time. They asserted that Korbak believed he was acting within policy by communicating with outside evaluators, and that Balesni coordinated his work with OpenAI board members and executives. Wang explained her dismissal was reportedly due to accessing an executive's email for recruiting purposes, a request she made to have removed which was not actioned.