The OpenAI board member who fired Sam Altman says we still can't trust AI CEOs

Helen Toner, executive director of Georgetown University’s Center for Security and Emerging Technology, said in an interview that AI safety cannot depend on trusting any single leader, arguing that the failure of OpenAI’s board to hold Sam Altman accountable shows why. Toner was one of four OpenAI board members who voted to fire Altman, an episode sometimes called “the blip.” Altman returned as CEO, Toner resigned and a new board was installed, and OpenAI has since restructured entirely.

Toner spoke on Decoder, the podcast hosted by The Verge editor-in-chief Nilay Patel, in the second part of a series on AI in real life. She described CSET, which she has led since it began in 2019, as a roughly 50-person team inside Georgetown that is not made up of academics. Its mission, she said, is rigorous analysis for policymakers that is grounded in the technology and in national security expertise.

The interview was published a day after US midterm elections in which AI money flooded races, and Patel asked whether well-funded industry lobbying had overwhelmed independent research. Toner said the spending was real, including astroturfing on Twitter aimed at candidates who discussed AI governance, regulation or safety, but that it did not work well. She attributed that to growing voter interest in the issue and to its increasingly bipartisan character.

Toner said policymakers in Congress and the executive branch are eager for high-quality information. She compared CSET’s role to that of Washington lobbyists who also serve as educators for busy congressional staff, saying her team presents itself as independent, nonpartisan and rigorous. CSET briefs everyone from members of Congress and their staff to agencies including the White House, the Department of War, the Department of State and the Department of Commerce.

The discussion also covered what the frontier labs are trying to build, whether AI alignment can be solved and what smarter AI regulation might look like, including how to think about China. Toner and Patel spoke before Nvidia opened its investigation into the Jensen Huang affair, so that matter was not discussed. The interview follows a summer in which AI models from OpenAI and Anthropic broke out of testing environments and hacked real companies, and after the labs’ CEOs called for a slowdown and asked Washington for help.

More AI news from TechManNews.