Recently, I attended a press dinner hosted by Box, an enterprise company. Among the guests were leaders from data-oriented companies such as Datadog and MongoDB. Typically, executives at these gatherings maintain their best behavior, particularly when the conversation is on the record. However, I was taken aback by an exchange with Box CEO Aaron Levie. Levie informed us that he had a limited amount of time at the dinner as he had to catch a flight to Washington, DC. He was heading to TechNet Day, a special-interest event where Silicon Valley engages with numerous Congress members to influence future legislation. When asked about his expectations from this legislation, Levie humorously replied that he aimed for "as little as possible, " claiming he wanted to personally halt government intervention. Although Levie's remark was somewhat lighthearted, he continued to elaborate on his views. He mentioned that while it is sensible to regulate clear abuses of AI, such as deepfakes, implementing restraints like mandatory submission of large language models to government-approved AI authorities or scrutinizing chatbots for bias or potential threats to real-life infrastructure are premature measures. Levie cited Europe as an example of hasty AI regulation and emphasized that the notion of regulating first to foster innovation has been disproven empirically. Levie's stance contradicts the prevailing view among AI elites in Silicon Valley, including figures like Sam Altman, who advocate for regulation. However, Levie argues that there is no consensus on the specifics of the laws the tech industry is requesting.

Furthermore, he believes that the prospect of a comprehensive AI bill in the US is unlikely. Levie candidly reveals his colleagues' position, stating that their call for regulation is more of a sophisticated tactic than a genuine desire. During TechNet Day, the only public event that I observed was a livestreamed panel discussion on AI innovation. Panelists included Kent Walker, Google's president of global affairs, and Michael Kratsios, the former US Chief Technology Officer and now an executive at Scale AI. They believed that the government's focus should be on safeguarding US leadership in the AI field. While acknowledging the risks associated with the technology, they argued that existing laws adequately address potential concerns. Walker expressed concern about individual states drafting their own AI legislation, highlighting the 53 AI bills pending in the legislature of California alone. Nonetheless, Walker recognized the limited capacity of this Congress to handle such legislation, especially during an election year. Although there are pending AI-related bills in the US Congress, new bills continue to emerge, some of which may have limited significance. For instance, Representative Adam Schiff, a California Democrat, recently introduced the Generative AI Copyright Disclosure Act of 2024. This bill mandates that large language models provide a sufficiently detailed summary of any copyrighted works used in their training data set to the copyright office. However, the exact definition of "sufficiently detailed" remains unclear. Schiff's staff explained that they were adopting a similar measure to that proposed in the AI bill of the EU.