Nat Purser
i agree that antitrust enforcers could do more under existing laws to challenge deceptive, unfair, and anticompetitive practices in ai markets now. but this doesn’t really answer the live question: how can competing labs coordinate to slow down the race dynamic she describes here, or share safety-relevant info? many of the harms described below — like “releasing unvetted ai models,” “shipping flawed ai tools,” and “models participating in criminal activity” — are about creating liability for harm *post* deployment. ex post liability for a harmful product is good, but an insufficient regulatory framework for a technology where some of the greatest risks lie in its development (see: Hugging Face hacks perpetuated by unreleased models!) the harms could be so severe that imposing liability afterwards is a deeply inadequate remedy. we need legal frameworks that let firms share info and coordinate in narrow, transparent ways (with government oversight) before irreversible, catastrophic harms occur.
Lina Khan
Law enforcers already have authority to charge companies and their CEOs for creating and releasing dangerous, unvetted, or defective products. We shouldn’t let discussions about new legal regimes distract from the fact that there’s no AI exemption from laws already on the books —