Washington: Former OpenAI safety employee David Robinson has criticised the company’s approach to artificial intelligence safety after resigning, arguing that a culture focused on rapid development could increase the risk of failures. His comments were published in The Atlantic on Saturday.
Robinson, who spent three-and-a-half years at OpenAI, said AI companies were not being “nearly careful enough” and should give greater priority to safety expertise and research before developing more capable systems.
He wrote that “the time for trial and error is over” and argued that advanced AI should have safeguards comparable to those used in high-risk sectors such as nuclear power and aviation.
He said OpenAI relies heavily on “iterative deployment”, in which systems are released and safeguards strengthened when problems emerge.
Robinson helped draft OpenAI’s Preparedness Framework and oversaw safety reports for 12 frontier-model launches. He said the company’s rapid launch cycle was preventing it from achieving the level of care he considered necessary.
OpenAI defended its safety practices, saying it ensures models do not become more capable than the company can safely manage and secure, and that training can be paused or models held back when necessary.
Robinson also warned that AI capabilities are advancing faster than researchers’ understanding of alignment, which focuses on ensuring AI systems act according to human goals and values.
The remarks add to a wider industry debate over the pace of AI development and safety safeguards.
