Uploaded August 2026 | Updated September 2026, 2 weeks ago
How worried should we be when an AI bot goes rogue?
BKC Faculty Director and SEAS computer science professor James Mickens breaks down the recent incident where two Open AI models broke through their test environment (also known as a “sandbox”) and independently hacked into Hugging Face’s servers. BKC spoke with Mickens to learn more about what the hack exposes about AI safety, and the work his research group is doing to help prevent these sorts of autonomous attacks in the future.
James Mickens is the Harvard John A. Paulson School of Engineering and Applied Sciences computer science professor and Faculty Director of the Berkman Klein Center for Internet & Society. Learn more here: https://cyber.harvard.edu/people/jmickens
How worried should we be when an AI bot goes rogue?
BKC Faculty Director and SEAS computer science professor James Mickens breaks down the recent incident where two Open AI models broke through their test environment (also known as a “sandbox”) and independently hacked into Hugging Face’s servers. BKC spoke with Mickens to learn more about what the hack exposes about AI safety, and the work his research group is doing to help prevent these sorts of autonomous attacks in the future.
James Mickens is the Harvard John A. Paulson School of Engineering and Applied Sciences computer science professor and Faculty Director of the Berkman Klein Center for Internet & Society. Learn more here: https://cyber.harvard.edu/people/jmickens
