![\"Anthropic](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/04\/anthropic-ai.jpg\")
<\/p>\n<\/p>\n
President Donald Trump<\/a> said thatartificial intelligence firm Anthropic<\/a> is “shaping up” for a “possible” deal to allow the Pentagon to use the company’s AI technology, signaling a potential reversal after the White House and the Defense Department<\/a> moved earlier this year to cut the company off from federal work.<\/p>\n Speaking on CNBC’s Squawk Box<\/a>, Trump said Anthropic officials had recently come to the White House for what he described as “very good talks.” He added that the company was “very smart” and “can be of great use.” When asked directly whether a Pentagon deal could be ahead, Trump answered, “It’s possible. We want the smartest people.”<\/p>\n The remarks suggest a possible thaw in one of the most closely watched standoffs between Washington and a leading AI developer. In February, Trump directed the government<\/a> to stop working with Anthropic, and the Pentagon later designated the company a supply-chain risk.<\/p>\n That decision followed a dispute over guardrails Anthropic wanted around military use of its systems, particularly concerns that its models could be used in ways that clashed with the company’s safety policies. Anthropic challenged the determination<\/a> in court in March.<\/p>\n The latest shift appears to have followed direct outreach from Anthropic’s leadership. CEO Dario Amodei met with White House officials last week in an effort to repair the relationship, with the White House describing that meeting<\/a> as “productive and constructive.” Anthropic, for its part, said the conversation focused on shared priorities, including cybersecurity, maintaining America’s lead in the global AI race, and AI safety.<\/p>\n The renewed dialogue comes as Anthropic has been drawing fresh attention in Washington because of its newest high-end model, Claude Mythos<\/a>, which has raised alarms and interest because of its potential ability to identify cybersecurity vulnerabilities and even devise ways to exploit them. Anthropic has said the model will not be made generally available. Instead, it launched Project Glasswing<\/a>, allowing selected companies and organizations to privately evaluate the technology and prepare defenses against the kinds of risks it may expose.<\/p>\n That backdrop helps explain why the administration may be reconsidering its posture. Even after the Pentagon blacklist, White House officials continued to engage with advanced AI labs about frontier models and software security. The Associated Press reported<\/a> that any technology under consideration for federal use would still need to undergo a technical evaluation period, suggesting that even if political relations improve, a formal deal is not imminent.<\/p>\n