What AI CEOs still don't get about Washington

• But Congress — on privacy, social media and now AI — has a history of getting stuck in the policy weeds, and lawmakers are now grappling with heavy lobbying and growing constituent demands on the future of the tech.

Catch up quick: OpenAI's new industrial policy paper describes AI changing the world on a scale similar to the Industrial Revolution, requiring aggressive policies like tax reform or a four-day workweek, while others have been floated by progressives for decades, such as boosting child care.

• Anthropic's policy ideas have skewed more toward internal governance and transparency, such as economic audits to determine AI's impact on jobs, along with stricter export controls and greater government evaluation of AI systems.

Behind the scenes: The architects of these proposals aren't new to the policy world.

• OpenAI chief global affairs officer Chris Lehane has long argued for redistributing the gains of new technologies, from pitching a "new deal" for crypto to promoting policies that would spread AI's economic gains more broadly.
• Anthropic meanwhile is ramping up its D.C. presence under public policy head Sarah Heck, who was previously at Stripe and worked on global entrepreneurship and public diplomacy at the White House National Security Council under former President Obama.

What they're saying: Lehane, a longtime political operative, told Axios that OpenAI is focused on promoting these policies at the state level, where there's a higher chance of success, especially in an election year when voters want to ensure they benefit from AI.

• "There is one truism in every campaign, which is, every politician says they lead, but what they typically do is they follow where the voters are, and they will move very quickly if they see voter sentiment on it," Lehane said.
• "We know the majority of Americans want government to take action on these issues," Heck said, pointing to Anthropic's policy positions on model transparency, economic impacts and energy.

Anthropic has also backed state-level AI transparency bills to mitigate the technology's biggest risks while calling for a federal standard, saying that transparency is the first step to give policymakers and the public visibility into how systems are developed.

The big picture: Silicon Valley and Washington are often speaking different languages: One moves fast and breaks things, while the other moves slowly — if at all.

• "Both coasts think that they're in charge," Nand Mulchandani, former chief technology officer of the CIA and of the Pentagon's Joint Artificial Intelligence Center, told Axios in an interview at the HumanX conference in San Francisco this week.
• "But Silicon Valley now has power rivaling the power of what a government has. What we're seeing now is the first large fight over who's driving the bus."

While the AI industry has allies in the White House, the Trump administration has also run into limits in Washington.

• Efforts to preempt state-level action, for example, have repeatedly failed, and the White House's most recent AI framework proposal for Congress faces an uphill climb.

The bottom line: AI companies can float sweeping policy ideas knowing they're unlikely to go anywhere, and still claim they warned Washington.