• Commerce Secretary Howard Lutnick's remarks signal a thaw in relations with Anthropic after a contentious legal battle.
  • A federal judge's ruling struck down the Pentagon's blacklist of the AI firm, citing retaliation for protected speech.
  • The dispute centered on AI safeguards for military use, raising questions about who sets limits on technology in national security.

A Shift in Tone

Commerce Secretary Howard Lutnick told attendees at the G20 Innovation Ministerial in Chapel Hill on September 2 that Anthropic is "back on the right side" of the Trump administration, according to Axios. The comment marks a sharp reversal from June, when the Pentagon designated the Claude-maker as a national-security "supply-chain risk" — a move typically reserved for foreign-linked entities.

Lutnick characterized the temporary controls on Anthropic's most powerful models as a one-off and said AI systems should be allowed to compete when "reasonably" guardrailed. His remarks came just days after U.S. District Judge Rita Lin struck down the designation on August 28, ruling that the administration had acted unlawfully in retaliation for Anthropic's policy positions.

The dispute began in February when negotiations over a Defense Department contract broke down. Anthropic refused to remove two safeguards from Claude's military-use terms: prohibitions on mass domestic surveillance and fully autonomous lethal weapons. The company said it otherwise supported lawful national-security applications. The Pentagon demanded "all lawful uses" access, but Anthropic wanted explicit binding restrictions.

"The court's decision is a significant victory for our ability to ensure our technology is used in ways that align with our values," a company spokesperson said in a statement. Anthropic co-founder Tom Brown was scheduled to appear at the same G20 event, though it's unclear if he and Lutnick met.

Anthropic, a public-benefit corporation founded in 2021 by former OpenAI leaders, has seen explosive growth. Reuters reported a $965 billion post-money valuation after a $65 billion funding round in May. The company's annualized revenue run rate surpassed $65 billion by the end of July, up from $47 billion in May. Claude Code alone had topped $2.5 billion in annualized revenue by February.

A Fragile Truce

Lutnick's comments imply the White House views the conflict as manageable rather than an enduring breach. However, this may not be a complete settlement. Reporting indicates a separate legal case in Washington remains ongoing, and the government could appeal the California ruling.

"This is a political reset, not the end of the policy dispute," said Sarah Kreps, a tech policy analyst at Cornell University. "The core question — whether AI companies or the government set enforceable boundaries for military and surveillance uses — remains unresolved."

The reversal reduces the risk that a policy disagreement could suddenly cut a major AI supplier off from government work. It may make private developers more willing to negotiate military contracts, while also making federal agencies more cautious about using supply-chain-risk powers for domestic companies.

Lutnick framed the broader policy goal as maintaining U.S. leadership over China in AI. The G20 meeting is about both industrial policy and diplomacy: preserving technological advantages, promoting U.S. firms abroad, and influencing global governance. The administration is pushing for relatively light-touch AI rules, a stance that could accelerate innovation but also intensify concerns over surveillance, labor disruption, and military autonomy.

Anthropic had a Defense Department contract with a ceiling of up to $200 million, but its much larger commercial business reduces dependence on federal revenue. The company's rapid revenue growth suggests businesses are moving beyond experimentation toward paying for coding, knowledge-work, and workflow automation tools. That supports cloud providers, chipmakers, and data-center construction, but may also pressure firms and workers to adopt AI-enabled productivity tools.

The episode will likely influence competitors. OpenAI reached a Pentagon agreement that it said incorporated protections related to domestic mass surveillance and human responsibility for force — similar themes to Anthropic's requested safeguards. This suggests the market may be converging on contractual and technical controls rather than unrestricted access.

Correction: An earlier version of this article misstated the date of Lutnick's remarks. They were made on September 2, not September 1.