Kanter: no antitrust exemption needed for AI safety

Kanter's argument lands in the middle of a widening debate about who controls the rules for frontier AI. He describes the industry's own safety push as ranging from a generous fear of driving with no lines on the road to a more cynical desire to pause spending and protect IPO valuations. In his telling, neither reading justifies an antitrust exemption. The friendly, existing path for cooperation is narrower: a repository where companies can share information about threats and malicious bots, which he says other industries already do without violating competition law.

His comments also highlight an unusual political alignment. He notes that David Sacks, a former Trump AI czar, has approvingly retweeted Lina Khan's view that existing antitrust law is sufficient, while figures from opposing camps have shared stages to call for a halt to AI development. Kanter traces part of this to the 15–20 years of hands-off treatment that began with Section 230 and lasted across administrations, which he says eroded public trust in tech companies.

Looking ahead, he places the immediate burden for clear liability rules on Congress, which he describes as unable to act, leaving product liability and state-level rules to fill the gap. The Meta child-safety case, which took a decade and ended in a settlement, is his example of how slow that backstop can be. For the AI labs, the practical test is whether they can show that coordination on safety stays within existing law, or whether their requests are read as an attempt to soften competition ahead of public listings.