A hypothetical AI model-release pact and the competition question
One participant suggests labs could hold back models with certain bio capabilities until they agree on a testing protocol.
TLDR
One participant asks whether cybersecurity regulation already covers the issue and argues that AI “pacing” is a different endeavor. In reply, another says the issue is broader than cyber: labs could agree not to release models with certain bio capabilities until they settle a mutual testing protocol. That participant says similar coordination in other industries, such as airline price coordination, would typically be considered anti-competitive.
Combined views
330
1 Source, first seen ago