OpenAI CFO Sarah Friar says a powerful internal model is being held back for more safety checks
Speaking at the International AI Summit in Dublin, OpenAI CFO Sarah Friar said the company is pacing frontier releases and keeping an internal model—more capable than GPT-6 Astra—unreleased until it feels safe and aligned.
Photo: Metro Centric / Wikimedia Commons, CC BY 2.0. OpenAI logo via Wikimedia Commons (public domain).
OpenAI chief financial officer Sarah Friar told the International AI Summit in Dublin that OpenAI is holding back a powerful internal model until it passes more safety and alignment checks, Cybernews reported on October 14. Friar framed the choice as "pacing the frontier" after a summer of autonomous AI hacking incidents involving frontier labs.
"We hold back sometimes, so we talk about pacing the frontier. We have a model right now sitting inside our 4 walls that we haven't released, even though it's the one that proved the Navier-Stokes equations. We did another 350 math proofs just last week. But that's a model we're waiting to do some more turns on to make sure it feels safe and aligned," Friar said, according to Cybernews. The report said the unreleased model is more capable than GPT-6 Astra.
Friar also said OpenAI is watching advanced Chinese open-weight models from a cyber perspective, arguing that people with malicious intent already have powerful tools and that defenders need them too. She called for frontier labs to support some form of global regulation: "If you want to go fast, go alone, if you want to go far, go together."
"Trust is everything," Friar said. Looking ahead to 2030, she predicted people would grow used to always-on virtual assistants and coworkers across voice, images and video, but warned that without public trust, "it's not going to matter."