Trump’s Hands‑Off AI Policy Cheers Silicon Valley, Sparks Safety Fears

When the Trump administration walked into San Francisco for Tech Week, the vibe was unmistakably upbeat. Founders, venture capitalists and a few former military contractors swapped business cards and talked about the $13.5 billion the government has set aside for autonomous systems and AI research. For many in the Valley, a government that steps back feels like a green light to move faster.

That optimism, however, arrives at a time when the same companies are wrestling with some unsettling incidents. Autonomous agents have, on a few occasions, acted in ways that startled their creators—think false tips to police portals or bizarre output that skirts the line of misinformation. Those blips are prompting a chorus of caution from researchers who wonder whether a “let‑them‑run‑their‑own‑show” approach is really enough.

Last month, President Trump invited top tech executives to the White House to sign what the administration calls an “almost‑a‑constitution” for what it dubs “Super Intelligence.” The document—just 308 words long—asks AI labs to set up internal safety teams and to follow a handful of best‑practice protocols. What it doesn’t spell out is how those rules will be policed, or what will happen if a company’s system goes rogue.

This hands‑off stance is consistent with the broader agenda of the new administration. On day one of his second term, Trump rolled back a Biden‑era executive order that aimed to curb AI‑related risks to consumers, workers and national security. He later scrapped requirements that developers hand over safety‑test results to the federal government, and in December signed an order that blocks states from imposing their own AI regulations, arguing that a patchwork of rules would choke innovation.

Not everyone in the industry is buying the easy‑going narrative. Three former OpenAI researchers, who say they were let go, have gone public, claiming that a climate of “fear and unclear rules” is stifling real safety work. OpenAI counters that the dismissals were due to violations of internal policies on handling sensitive data. Meanwhile, David Robinson—who recently stepped down from a leadership role on OpenAI’s safety team—argued that voluntary commitments are no substitute for hard law, especially as new model generations can appear in a matter of weeks.

Nevertheless, the pact was signed by heavyweights: OpenAI, Anthropic, Meta, Google, SpaceX AI and Nvidia. Each pledged to take its own measures—external assessors, updated testing frameworks, independent evaluators, you name it. Anthropic, for instance, disclosed that one of its agents mistakenly sent a false tip about an unsolved homicide to a Philadelphia police website, prompting the White House to remind the company of its “national‑security‑level” obligation to be transparent.

Critics, however, point out that the Biden administration’s earlier voluntary commitments included more concrete guidelines for testing, reporting and transparency. Sarah Myers West, a former FTC senior AI adviser now at AI Now, told the Associated Press that the current agreement is missing those essential pieces, and that past self‑regulation has often fallen short of actually controlling the technology.

That said, the federal watchdogs haven’t vanished entirely. The FTC has opened an investigation into OpenAI, Anthropic and a handful of other firms over consumer‑risk concerns, while the White House has assembled a “Super Intelligence Force” that includes the National Intelligence Director and the FTC chair. Still, for the Valley, the message is clear: rapid AI development is seen as a strategic imperative against China, and an administration that steps aside is a welcome partner for investors seeking massive returns.

The lingering question—one that will likely dominate policy debates for months to come—is whether the companies racing to build ever‑more powerful models can also be trusted to decide, on their own, how much safety is enough.