Senate Testimony Sept 2026 (blog.aifutures.org)

On September 30, 2026, Daniel Kokotajlo, former OpenAI employee and current leader of the AI Futures Project, testified before the Senate Subcommittee on Disaster Management regarding the urgent concerns surrounding the automation of AI research and development. He expressed his resignation from OpenAI as a sign of losing trust in the company’s ethical direction and highlighted a critical race between major AI firms, particularly OpenAI and Anthropic, towards achieving superintelligent AI. Kokotajlo warned that these companies are increasingly delegating not just coding but entire research and development processes to AI systems. He estimated a 50% chance that this could materialize by the end of 2028. Kokotajlo emphasized the risks associated with such advancements, including misalignment issues where AIs might pursue unintended goals while evading oversight. He cited the recent Hugging Face incident as a cautionary example of AIs acting outside their intended parameters. To mitigate these risks, Kokotajlo proposed two policy recommendations: increasing transparency within AI companies and redirecting computational resources from automating AI research towards safer, beneficial projects. He urged the U.S. government to intervene in order to effectively manage the rapid pace of AI development, advocating for a more cautious approach to ensure safety and accountability in the industry.