Trump Signs Voluntary AI Safety Accord at White House Summit

President Donald Trump signed an executive order Tuesday directing federal agencies to replace references to artificial intelligence with the term super intelligence, a move that reframes ongoing debates over the technology's risks while leaving regulatory oversight in the hands of the companies developing it. The order followed a White House luncheon attended by executives from OpenAI, Anthropic, Alphabet, Meta, Nvidia, and other firms, including Elon Musk and Dario Amodei.

The renaming applies immediately to government communications and programs that already incorporate the technology across agencies. Trump described the change as a way to move past what he called overstated fears, telling reporters that the previous term had carried connotations of a hoax. He presented the rebranding as a fresh start that better reflected the technology's potential for American innovation and economic growth.

Alongside the order, participants signed a two-page document titled the White House Accord on Super Intelligence. It commits each company to developing its systems safely and in ways that maintain public trust, but contains no enforcement mechanisms or independent oversight provisions. Trump characterized the agreement as morally binding rather than legally required. This approach aligns with the administration's consistent resistance to new regulatory structures, even as several signatories have previously called for stronger government guardrails on advanced models.

The event highlighted tensions within the industry itself. Amodei and other executives have warned in recent weeks about the need for policy responses to potential extreme risks from frontier systems. Their presence at the White House underscored the gap between those public statements and the voluntary framework now in place. No new standards for testing, transparency, or deployment timelines emerged from the gathering.

Federal employees will now navigate updated terminology in official documents and procurement processes, adding administrative steps without altering underlying technical capabilities or safety practices. The broader public will likely continue using the original term in everyday discussion, limiting the order's practical reach outside government circles.

The episode reflects a broader pattern in technology policy where symbolic actions substitute for institutional constraints. By emphasizing industry self-policing and rhetorical shifts, the administration has preserved existing power arrangements between government and leading technology firms. Questions remain about how this stance will interact with state-level efforts and international regulatory developments that continue to advance separate standards for advanced AI systems.