Sen. Adam Schiff (D-CA) used a Decoder podcast interview to criticize the Trump administration's approach to AI safety, calling the recent White House signing of a non-binding pledge by tech executives "very jarring." Schiff noted that many of the same leaders privately say they want real regulation, but publicly go along with a voluntary pact to avoid angering the president. He argued that the pledge does nothing to address the serious risks posed by increasingly advanced AI systems.
Schiff pointed to specific national security dangers, including the unbridled use of AI at the Pentagon and the potential role of AI in target selection during the bombing of an Iranian girls' school. He also cited the administration's efforts to go after Anthropic for refusing to support domestic mass surveillance and fully autonomous weapons. In his view, these are issues that require binding oversight, not promises.
On the regulatory path forward, Schiff said he supports creating a new agency to oversee AI. But he acknowledged a major legal hurdle: the Supreme Court's Loper Bright decision ended Chevron deference, meaning courts—not agency experts—will resolve ambiguities in any new law. That could leave even well-designed AI rules tied up in litigation for years, making the difference between a real safety regime and a symbolic one. The interview underscores how AI policy has become entangled with broader political battles over executive power and corporate accountability.