U.S. Drafts Strict AI Rules After Clash With Anthropic
Synopsis
The Trump administration has proposed new rules for A.I. companies seeking government contracts. New York’s rules state that any AI firm working with the US government must grant it an irrevocable licence to use its AI tools for whatever legal purpose it likes. No limits. No red lines. The action comes immediately on the heels of a meeting with Anthropic at the Pentagon, which declined to allow military use of Claude without restrictions around surveillance and autonomous weapons. The new rules would ensure that no company could ever again do what Anthropic did.
The Trump administration is now drafting rules that would require any AI company seeking US government work to grant the government a permanent, no-strings license to use its AI for any legal purpose. The rules are directly descended from the Anthropic standoff, in which Anthropic had refused to let the Pentagon use Claude unconditionally and was cut off for it.
Key Highlights
- U.S. government drawing up rules mandating A.I. companies to permit all lawful uses of their tools, no holds barred
- Any AI company seeking a government contract is required to provide an indefinite, unfettered license
- The rules are a direct response to Anthropic’s refusal to provide the Pentagon unfettered access to Claude
- OpenAI did a deal with the Pentagon soon after Anthropic was frozen out, filling its space on classified networks
The U.S. Is Making New Rules So No A.I. Company Can Ever Again Do What Anthropic Did
The Trump administration has proposed new requirements for AI companies hoping to do business with the US government. The rules are simple, any artificial intelligence company that hopes to win a government contract must agree to allow the government to use its tools for whatever purpose is legal. No restrictions. No conditions. No red lines. The details were reported on Friday after it saw a draft of the new guidelines.
The regulations are a direct outgrowth of that battle. The government never wants any A.I. company to tell it what it can and can’t do with the tools that it pays for.
What the New Rules Really Mean
The draft states that businesses making use of AI when bidding for government contracts must provide what is known as an irrevocable licence. That represents a kind of permanent permission slip, one that cannot be rescinded, for the US government to use the AI system for any purpose allowed by law. A company had no business coming in down the line and saying certain usages were off limits. Once you sign, that is it.
Why this matters is because this is exactly what Anthropic attempted to do. It wanted contractual provisions prohibiting the Pentagon from using Claude for mass surveillance on Americans or for weapons that fire automatically without a human making the final decision. The government rejected both conditions, deemed them a national security threat, and ultimately severed all ties with Anthropic.
The Fight That Started This
Anthropic and the Pentagon had been negotiating in private for months. Non-negotiables, Anthropic’s chief executive, Dario Amodei, stood firm on, no mass domestic surveillance, no fully autonomous weapons. The Pentagon said such limits were unworkable. It wanted to be able to use any artificial intelligence it paid for however the law allowed. Officials said it was unacceptable and dangerous for a private company to impose conditions on military tools.
When negotiations fell apart in late February, Trump took to social media and ordered each federal agency to cease immediately using Anthropic. The defence secretary, Pete Hegseth, followed up by officially classifying Anthropic as a supply chain risk, an offence reserved until then for foreign adversaries like Huawei and Kaspersky. The move was not legally sound, Anthropic said, and it announced plans to take the government to court.
OpenAI Moved In, Then Hit Its Own Snags
Hours after Anthropic was cut off, OpenAI announced it had secured a new deal with the Pentagon to take Anthropic’s place on classified military networks. The swift nature of it raised eyebrows instantly. OpenAI said its own agreement had also included measures against domestic surveillance and autonomous weapons, the very issues that Anthropic raised. But the company soon was forced to reverse aspects of its pact amid criticism that it had acted too quickly.
CEO Sam Altman publicly acknowledged that the deal appeared hasty and botched. OpenAI changed the language in the contract. The past few days were a scramble illustrating how messy these government A.I. deals can be, even for a company trying to do the right thing on deadline.
FAQs
- What are the new US AI rules about?
Any AI company that collaborates with the government must permit all legal applications for its tools, with no limitations whatsoever.
- Why are these rules being made now?
Because Anthropic wouldn’t allow the Pentagon unlimited access to Claude, and the government doesn’t want it to happen again.
- What is an irrevocable licence?
An unlimited license that can be revoked, once given, the government has free rein to do with the AI as it sees fit within the law.
- Did OpenAI replace Anthropic?
Yes, it forged a Pentagon deal hours after Anthropic was barred, though it was forced to amend aspects of the contract on the fly in response to criticism.
- Is Anthropic taking legal action?
Yes, the government’s actions are legally unsound, and the company will fight them in court, according to its chief executive officer, Dario Amodei.
Follow Inspirepreneur Magazine for the business news.
At Inspirepreneurs Magazine, covering entrepreneurship, business failures, and the human stories behind the world's most ambitious founders. She writes at the intersection of strategy and storytelling.
