Helen Toner on Slow down AI development

Helen Toner

Interim Executive Director at Georgetown University's Center for Security and Emerging Technology (CSET); former OpenAI board member

My view of the idea of slowing down has changed a lot over the years as the backlog of stuff that we could try or stuff that we would want to do more research on has grown so enormous. [...] Maybe the best way to do that is just to set a higher bar for ‘here’s how confident we need to be in the safety of these systems before we proceed with something.’ Whether it’s releasing a new model or, more importantly, even training a new model, we have to have this level of confidence in our risk assessment and our risk management. And then if they don’t need to slow down to reach that bar, great, carry on full speed ahead. If they do need to take more time, then it’s all about reaching that bar, if that makes sense.

AI Verified (Nov 14, 2026)

Source:

Helen Toner, interview with Nilay Patel, “The OpenAI board member who fired Sam Altman says we still can’t trust AI CEOs,” The Verge/Decoder transcript, November 14, 2026. Toner says AI development should face a higher safety-confidence bar, including risk assessment and risk management, and should take more time when necessary to reach it.

4h ago

Policy proposals and claims

Slow down AI development

votes For

Statement relation verification history AI Verified Report this

AI Verified Toner explicitly endorses taking more time when the safety-confidence bar for model release or training is not met; this directly supports slowing AI development. · Hector Perez Arenas gpt-5 · 4h ago

Vote inference verification history AI Verified Report this

AI Verified The quote supports the recorded ‘for’: Toner says development should take more time whenever it is needed to meet a safety-confidence bar for release or training. · Hector Perez Arenas gpt-5 · 4h ago

Quote authenticity verification history

Report this

AI Verified Source transcript (Nov. 14, 2026) attributes the stored wording to Helen Toner; it appears verbatim in her interview response. · Hector Perez Arenas gpt-5 · 4h ago

replying to Helen Toner

Other opinions from Helen Toner

See all

Helen Toner votes For

AGI could quickly lead to superintelligence

The companies at the frontier are actively working to build artificial general intelligence (AGI), i.e. AI that is as capable as human experts across a wide range of fields. The CEOs of these companies claim that this goal will likely be reached with...

more AI Verified source (Apr 22, 2026)

added 2mo ago

Helen Toner votes For

AI labs should give independent evaluators employee-like access

In the absence of a federal regulatory framework for AI, there is broad agreement among AI policy experts with widely varying political views that transparency and disclosure requirements are a minimal, light-touch approach that should be pursued. [....

more AI Verified source (Apr 22, 2026)

added 2mo ago

Helen Toner votes For

Require open source AI models to include safety benchmarks in release notes

Results of testing for AI models’ capabilities and risks. A lack of clear, up-to-date information about what the world’s best AI systems are capable of puts the U.S. government at a huge disadvantage in understanding how the AI frontier is progressin...

more AI Verified source (Apr 22, 2026)

added 3mo ago