Internally Deployed AI Models: Implications, Risks, and Gaps
This briefing has been postponed and will now take place on Wednesday, September 9, from 10:00–10:45 AM ET. We hope you’ll be able to join us at the new time!
The recent OpenAI/Hugging Face incident, involving a model that was never meant for public release, has highlighted the risks of internally deployed AI models. Alongside other recent incidents of unexpected AI agent behavior, including sophisticated social engineering campaigns and cyber breaches of third parties, it raises a hard question: What does responsible governance look like for models that were never supposed to leave the lab?
On Wednesday, September 9, from 10:00-10:45 ET, the Institute for AI Policy and Strategy (IAPS) invites you to a virtual briefing on the governance of internally deployed AI models.
This briefing will cover:
The risks posed by internally deployed models.
Why we should expect incidents like those we have seen over the past month to continue.
The current approach within frontier AI companies to securing internally deployed models.
Existing governance approaches and proposed federal legislation.
Policy recommendations.
There will also be an opportunity for Q&A.
The Institute for AI Policy and Strategy (IAPS) is a nonpartisan think tank that produces policy research to address the implications of AI, from today’s most advanced models to potential AGI and superintelligence. Our work equips policymakers and industry leaders to protect innovation while navigating high-magnitude risks and opportunities at the intersection of AI, national security, and geopolitics.