How Silicon Valley brought its AI fight to Washington

FOR most of the last decade, Silicon Valley treated Washington as an afterthought – a place to send lobbyists when a subpoena arrived, not a location to build strategy around. 

That posture has collapsed. In 2026, the arguments happening inside artificial intelligence labs and venture firms – speed versus safety, open models versus closed ones, how much deference the industry deserves – are no longer confined to Sand Hill Road.

They are being fought, on the floor of the US Congress, in the West Wing and in state capitals across the country.

The China narrative

The most consequential export from Silicon Valley to Washington is a story: that the US is locked in an existential AI race with China, and that anything slowing American labs down, such as regulation, export controls and liability, amounts to unilateral disarmament.

But some critics, including AI researchers, argue this narrative rests on a distorted picture of Chinese AI priorities and has crowded out more careful policymaking on both sides of the Pacific.

Whatever its accuracy, the framing has proven durable. Attaching “China” to any policy proposal in Washington has become close to a rhetorical master key. It shows up in fights over chip export rules, in the Pentagon’s growing appetite for frontier AI partnerships and now in a new fault line: distillation.

That fissure surfaced dramatically in July, when Chinese lab Moonshot AI released a model competitive with the best American systems, built in part – as Anthropic has argued – by “distilling” capabilities out of US models through mass automated querying. 

Anthropic went public in February describing an operation it said used over 24,000 fake accounts to extract Claude’s knowledge at industrial scale, and framed the practice as a national-security problem requiring urgent controls.

But Silicon Valley is not speaking with one voice here, and neither is Washington. 

A July letter backed by signatories including Nvidia, Microsoft, Meta, IBM, Dell and Palantir pushed in the opposite direction, warning that broad restrictions on open-weight Chinese models would reduce competition and entrench the largest proprietary vendors. 

While the letter did not name those companies, it was implied to mean companies such as Anthropic and OpenAI. 

That is a genuine industry rift, and it means lawmakers weighing export and distillation rules can no longer treat “the tech industry” as a single lobbying bloc with one ask.

Who gets to write the rules?

If the China frame supplies the rhetoric, the fight over federal pre-emption of state AI laws supplies the mechanics. 

States have not waited for Washington: By mid-2026 they had already enacted well over a hundred AI-related laws for the year, with Colorado’s Act, covering “consequential” automated decisions in housing, employment, healthcare and education, as the major example.

The industry has pushed hard for Congress or the White House to override that state-led patchwork with a single national standard.

The Trump administration has obliged, first through failed attempts to pass a 10-year moratorium on state AI enforcement, and more recently through a March 2026 executive framework calling on Congress to pre-empt “cumbersome state AI laws” while carving out exceptions for areas such as procurement. 

The politics here runs contrary to the usual partisan map: Republican Senator Josh Hawley voted against the original moratorium and has pushed instead for federal chatbot-safety legislation aimed at protecting children, while Ron DeSantis, Florida’s Republican governor, has publicly called pre-emption a giveaway to Big Tech. 

Congressional Democrats, meanwhile, have introduced legislation specifically to block federal pre-emption of state authority. 

The result is a debate less about whether AI should be regulated than about which level of government should be in charge.

Money finds a more direct route

Where the industry used to rely on lobbying shops to move opinion inside political institutions, 2026 has brought something more direct: campaign spending designed to shape who sits inside the establishment.

A new political action committees network called Leading the Future has pledged more than US$100 million to back AI-friendly candidates and oppose those seen as hostile to the industry, ahead of the 2026 midterms.

It is backed by venture capital fund Andreessen Horowitz, OpenAI president Greg Brockman, Palantir co-founder Joe Lonsdale and other prominent AI investors.

Supporters frame this as a defence of “sensible guard rails” against a state-by-state regulatory patchwork; critics see an attempt to buy influence over the pre-emption and safety fights before they are even decided. Either way, it marks a shift from persuasion to electoral leverage.

None of this suggests Silicon Valley is winning outright. More than 40 state attorneys-general have jointly warned AI companies they intend to use the full weight of their authority to address chatbot harms to children. 

And the industry’s own internal splits, on distillation and open-weight models especially, mean Washington now has competing factions of Silicon Valley to listen to rather than one unified voice.

What has changed is the posture of the tech industry. It once treated the government as either irrelevant or an obstacle to be managed from a distance. 

Now, the industry has concluded that it needs to be inside the room, funding campaigns and negotiating executive orders, rather than opposing regulation in the abstract. 

Washington, for its part, has largely adopted Silicon Valley’s own framing of the stakes: national competitiveness, the China race, the fear of over-regulating a technology still finding its shape.

The debate has not moved from one coast to the other so much as it has become one continuous argument, conducted simultaneously in both places, with the outcome in each feeding directly into the other.