In a recent interview on The Verge’s Decoder podcast, Jonathan Kanter, former head of the Antitrust Division at the U.S. Department of Justice under the Biden administration, addressed the growing calls from leading artificial intelligence firms for regulatory exemptions. Currently serving as a professor of law at Washington University in St. Louis and a professor of technology policy at Carnegie Mellon University, Kanter brought extensive experience prosecuting major tech cases against Google, Apple, and Ticketmaster.
The discussion centered on a widening debate regarding AI safety, with researchers from prominent labs like Anthropic and Google DeepMind resigning in protest over perceived negligence. Some experts estimate the risk of catastrophic AI outcomes at greater than 10 percent, prompting CEOs across the industry to advocate for slowed development and new regulatory frameworks, including requests for antitrust relief to facilitate coordination.
Kanter expressed skepticism toward these requests, noting that companies such as OpenAI, Anthropic, Google DeepMind, and Elon Musk’s ventures are fierce competitors who do not inherently trust one another. He suggested that their joint pleas for a slowdown may stem from a “prisoner’s dilemma,” where firms seek government enforcement of safety standards because they cannot rely on mutual restraint.
Offering a cynical counterpoint, Kanter proposed that these corporations might also be motivated by financial pressures. He noted that many firms are “hemorrhaging cash” and may fear that valuations could suffer ahead of initial public offerings if they do not appear to manage their pace of innovation responsibly. However, he firmly rejected the notion that these entities require an antitrust exemption to collaborate on safety.
“These companies do not need to coordinate in order to deliver safe and secure products to society,” Kanter stated, drawing parallels to the aerospace and automotive industries. He argued that manufacturers are responsible for ensuring their own products are safe without needing to collude with competitors to limit speed or features. He emphasized that liability should rest on the company whose digital agents cause harm, similar to how employers are held accountable for the actions of human employees.
While acknowledging that certain collaborative efforts, such as threat-sharing repositories, are permissible under existing antitrust laws, Kanter criticized the current regulatory vacuum. He likened the state of AI governance to having invented cars and trucks without establishing traffic laws, road markings, or speed limits. He pointed to legislative gridlock, influenced by factors such as gerrymandering and Citizens United, as a primary barrier to implementing necessary safety standards.
Kanter concluded that a lack of congressional action does not absolve technology companies of their duty to build secure systems. He asserted that addressing safety challenges is itself a form of innovation, and that the industry should focus on incentivizing the creation of safer, more reliable products rather than seeking exemptions to reduce competitive pressure.
Leave a Reply