Big AI Escalates Legal Fights While Buying Political Clout

Anthropic is simultaneously suing the U.S. government and briefing senior officials on its newest AI model, a split‑screen that shows how Big AI is contesting the very rules meant to govern it. The Claude maker has filed dual lawsuits challenging the Pentagon’s March decision to label it a national‑security “supply chain risk,” a designation usually reserved for foreign adversaries and one that effectively bans its systems from most defense work, according to reporting from TechCrunch and others. Yet days ago, CEO Dario Amodei was in the White House, where chief of staff Susie Wiles pressed him on Mythos, Anthropic’s new model, and its implications for national security and the broader economy, the Associated Press reported.

The stakes of that legal fight are enormous. A supply‑chain‑risk label can cost Anthropic hundreds of millions in government and contractor business and has already spooked commercial customers wary of doing business with a company framed as a national‑security problem, according to legal analyses cited by Mayer Brown and others. The company argues the Pentagon is punishing it for refusing to open its AI to “all lawful uses,” including autonomous weapons and mass surveillance, and for speaking out about that stance, a claim laid out in its complaints summarized by Forbes. A federal judge has already temporarily blocked enforcement of the designation while the case proceeds, calling the government’s approach an “Orwellian notion,” according to the AP.

At the state level, the same companies are reshaping liability rules. OpenAI is backing Illinois Senate Bill 3444, which would shield “frontier” AI developers from civil liability for catastrophic harms—such as AI‑enabled mass casualties or more than $1 billion in damages—so long as labs publish their own safety frameworks and avoid intentional or reckless misconduct, as detailed by WIRED and policy outlet PPC Land. Anthropic has quietly lobbied to weaken or kill the bill, warning it risks becoming a “get‑out‑of‑jail‑free card” for model makers if their tools are later used to build bioweapons or mount systemic cyberattacks.

Behind these legal battles, money is flowing to shape who sits in judgment. A new super PAC network, Leading the Future, launched with more than $100 million from Silicon Valley investors and AI executives to back candidates favoring rapid AI deployment and lighter-touch regulation, according to Wikipedia’s summary of the group drawing on FEC filings and press reports. Its affiliated nonprofit, Build American AI, can fund issue ads and dark‑money campaigns targeting skeptics without disclosing donors. Taken together, the lawsuits, bespoke liability shields and aggressive political spending reveal a core reality of the AI boom: the companies building powerful models are also racing to write, bend or buy the rules that will decide who pays when things go wrong.