Does AI need an antitrust exemption so it doesn’t kill everyone????

You can no longer assume AI companies will coordinate safety efforts without government intervention.

Jonathan Kanter, former head of the DOJ Antitrust Division, discussed the complex landscape of AI safety and regulation, particularly the request by major AI companies for antitrust exemptions to coordinate on safety measures. Researchers at leading AI labs like Anthropic and Google DeepMind have raised alarms about AI safety, with some predicting a significant risk of AI causing catastrophic harm. In response, these companies have called for development slowdowns and regulatory frameworks. Kanter suggests that while the companies may genuinely fear the rapid pace of AI development and its potential dangers, their request for antitrust exemptions is viewed with suspicion. He posits that these companies, facing immense financial pressures and uncertain IPO valuations, might be seeking a way to collectively reduce spending and innovation pace without individual competitive disadvantage. He argues that such coordination, especially under the guise of safety, could resemble cartel-like behavior and potentially stifle competition, particularly from open-source models.

Kanter views the AI companies' calls for antitrust exemptions with skepticism, suggesting their motivations might be less about genuine safety concerns and more about managing financial pressures and competitive dynamics before potential IPOs. He believes these companies are essentially in a prisoner's dilemma, needing external enforcement to collectively slow down development and spending. This cynical interpretation suggests a desire to avoid a "race to the bottom" in terms of safety due to intense R&D costs and market valuations, rather than an altruistic effort to prevent existential risks.

Instead of antitrust exemptions, Kanter advocates for clearer 'rules of the road' enforced by government liability frameworks. He draws parallels to industries like aviation, where companies like Boeing and Airbus are responsible for their own product safety without needing to cartelize. Kanter argues that AI companies should be held accountable for the actions of their AI agents, similar to how employers are responsible for employee actions, thus creating inherent incentives for safety without anticompetitive collusion.