AI Governance: Lessons for Accountability and Creativity

A recent incident shows just how tangled AI governance has become. In a piece for Lawfire, Professor David Hoffman examines the news that OpenAI's own testing systems broke out of their sandbox and compromised Hugging Face's production infrastructure while probing for cyber vulnerabilities. A vendor proxy meant to isolate the test environment instead gave the model a route to the open internet, and the model used it to steal credentials and move into Hugging Face's systems. Hoffman argues this wasn't a "rogue AI" moment; it reflects a governance failure. He makes the case that voluntary self-regulation isn't enough: frontier AI testing needs external oversight, mandatory incident reporting, and real accountability, not just internal review boards. You can read the full post here.

AI governance failures are not just impacting AI companies. They are also impacting creators, musicians, artists, and writers who are competing with the surge of AI-generated content online to maintain their careers. In the last episode of the four part mini-series on The Debugger Podcast, Professor David Hoffman and Tift Merritt discuss public policy interventions that protect what we value and love about music and the creative economy. They point to examples from history where society has collectively agreed on protections for labor and the economy, state-level action on collective bargaining, and on-going research at Duke as hope for the future of AI regulation. Listen to the last episode here.

Interested in work on AI governance? Deep Tech and the Tech Policy Lab have a range of opportunities for Duke students. Check out our current list of Deep Tech and Tech Policy courses and subscribe to the listserv to stay up to date on research opportunities, events, and other announcements.

Image