Dario Amodei, CEO of the artificial intelligence company Anthropic, has advocated for allowing AI firms to self-regulate to prevent potentially catastrophic outcomes. He has proposed granting these companies an antitrust exemption to coordinate efforts in slowing AI model deployment and encouraging similar controls in China. This approach gained visibility recently when former President Donald Trump, after a private White House lunch with top AI executives, expressed support for “tremendous self-regulation” within the sector.
However, critics argue that relying on AI companies to police themselves echoes past failures in technology governance. The experience of Facebook, now Meta, is often cited as an example. Around 2018, Meta introduced self-governance measures meant to oversee its operations, but these were seen largely as efforts to forestall government intervention. Subsequent legal proceedings revealed that Meta concealed evidence of harm caused to young users by its platforms, raising concerns about the effectiveness and sincerity of industry-led oversight.
Lina M. Khan, chair of the Federal Trade Commission from 2021 to 2025, has emphasized that the United States has a history of successfully managing transformative and potentially hazardous technologies through democratic governance. She points to regulatory frameworks developed for industries ranging from railroads to nuclear energy as precedents for how AI can and should be overseen. Khan contends that existing laws—such as product liability, tort law, and consumer protection statutes—already apply to AI companies, addressing defective or dangerous products and holding businesses accountable for negligence or deception.
State attorneys general have a role to play as well. While federal enforcement has sometimes been limited, state-level interventions, such as those pursued by Florida, illustrate mechanisms for halting harmful corporate practices and imposing penalties. Legal experts note that individual AI executives could also face personal liability, especially in cases involving unlawful activities or security breaches like the recent hack of Hugging Face.
Khan further argues that comprehensive legislative action is necessary to address AI’s unique risks. Drawing on regulatory models for banks, pharmaceuticals, and nuclear materials, Congress could introduce mandatory testing of advanced AI systems, break up dominant firms to prevent conflicts of interest, and establish truly independent oversight bodies. Legislative proposals have surfaced that would increase accountability, including measures to impose stronger criminal penalties on executives and even contemplate a “corporate death penalty” for egregious misconduct.
Underlying these challenges is a broader concern about the capacity of U.S. governance structures. Since the Republican takeover of Congress in 1995, with consequential cuts to committee staffing and internal expertise, legislative effectiveness has been diminished. Contributions from political donors and judicial trends favoring economic libertarianism have further complicated efforts to regulate corporate behavior. The Trump administration’s approach to enforcement has also been criticized for allowing wealthy interests to act with relative impunity.
Against this backdrop, critics warn that surrendering regulatory authority to private AI companies risks repeating past mistakes. The swift development pace and financial motivations of these firms have contributed to present challenges. Legal experts and policymakers urge that AI technologies should serve public interests under democratic oversight, not be left to the discretion of the very corporations that profit from their rapid deployment. They advocate for robust congressional investigation, swift legislation, and vigorous enforcement to ensure AI’s benefits do not come at unacceptable social costs.