According to people familiar with the matter, Anthropic, an artificial intelligence company that was once included on the national security blacklist, is easing tense relations with the Trump administration, and related interactions are showing signs of recovery as the company prepares for its IPO.
Earlier this year, Anthropic's relationship deteriorated sharply after it refused to allow the U.S. military to use its models for domestic surveillance and fully autonomous weapons systems. The U.S. government immediately placed the company on a national security blacklist. Relevant measures are expected to take effect later this year.
In March this year, the U.S. Department of Defense identified a domestic company as a "supply chain risk" for the first time. In the past, this label was mainly used for companies related to "rival countries."
This designation means that in the future, tens of thousands of contractors who provide services to the US military will not be allowed to use Anthropic's artificial intelligence products when involving defense projects.
People familiar with the matter said that relations between the two parties have shown signs of improvement since Anthropic CEO Dario Amodei went to the White House in mid-April to talk with officials about cooperation, although the company is still challenging the "supply chain risk" determination in court.
Franklin Turner, a lawyer in the field of government procurement and contracts, pointed out that the Department of Defense is still "actively" responding to the lawsuit and believes that as long as this dispute is not resolved, the broader harm to Anthropic's business prospects will be difficult to eliminate in the short term.
Turner said that once the government sends a signal that it is "disengaging" from a company, it is a "significant issue" for the company.
Observers believe that a symbolic but well-received "thaw" signal is that the White House invited Amodei to attend the signing ceremony of the artificial intelligence executive order originally scheduled to be held on May 21.
The event was ultimately canceled because Trump was unhappy with some of the terms, two people familiar with the matter said.
On Tuesday, Trump signed an executive order aimed at promoting innovation and safety in advanced artificial intelligence. Anthropic later issued a statement on the social platform X, saying it looked forward to "cooperating" with the White House in implementing the order.
A White House spokesman did not comment on the administration's relationship with Anthropic, and Anthropic's public relations team declined to comment.
In addition to the White House, Anthropic has recently communicated with a number of senior government officials about its most advanced system "Mythos", focusing on how critical infrastructure can respond to the risk of cyberattacks driven by artificial intelligence.
According to people familiar with the matter, Anthropic warned that Mythos has the potential to significantly enhance the capabilities and destructive power of such cyber attacks.
U.S. authorities collectively refer to important industries such as financial institutions, emergency services, and hospitals as critical infrastructure and believe that these areas require special protection to prevent attacks.
As interactions between all parties heat up, Anthropic is also stepping up its planned stock market listing.
Previous reports stated that the artificial intelligence giant’s IPO valuation is expected to reach US$1 trillion.
Market analysts pointed out that CEOs of large companies often take the initiative to strengthen relations with the White House during major policy and regulatory stages to avoid incurring the president's dissatisfaction; in Anthropic's case, closer government relations are also expected to boost investor confidence in the short term.
Harrison Rolfes, senior research analyst at venture capital data firm PitchBook, believes that the dispute with the government is more like "a short-term bruise" on the company's development path.
Another U.S. official said Anthropic employees met with Treasury Secretary Scott Bessant this spring to discuss Mythos and possible actions the president could take on artificial intelligence.
The official said the above talks provided policy reference for the executive order signed on June 2, which included requiring leading developers to submit their most advanced models to the government for network security testing and other purposes.
Although relations with the White House and Treasury Department have warmed, tensions between Anthropic and the Pentagon have not completely dissipated.
As of Thursday, the two sides are still submitting legal documents to the court on the issue of identifying "supply chain risks", and are fighting and defending each other.
At the same time, according to US military spokesman Sean Minton, Anthropic employees did not participate in the artificial intelligence cyber attack simulation exercise led by the Army on April 27.
The exercise attracted cybersecurity executives from several leading artificial intelligence developers, including representatives from companies such as Google and OpenAI.