Pentagon leadership is pushing back hard against high-profile warnings about artificial intelligence wiping out humanity, while simultaneously accelerating a migration of its most sensitive AI work away from Anthropic.

Emil Michael, the U.S. Defense Department's Chief Technology Officer, publicly rebuked recent warnings at a defense industry gathering in Washington on Thursday, directly addressing a widely circulated social media thread from a former Anthropic engineer. The post, which has amassed over 150 million views, alleged that insiders within AI companies believe the technology could "kill us all before the end of this decade."

Michael dismissed such concerns as a "doom loop," framing them as little more than "fear of data centers, fear of change."

Alongside the rhetorical rebuttal, Michael disclosed that the Defense Department has already shifted approximately 90% of its classified AI workloads off Anthropic platforms to alternative vendors, with the remainder slated for completion by the end of this month.

The CTO also characterized the ongoing legal battle between the Pentagon and Anthropic as a dispute with national security implications at its core. He pointed to a pending case before the D.C. Circuit Court of Appeals as the "more significant" arena, asserting that Anthropic continues to pose a supply chain risk to the U.S. absent a temporary injunction.

Viral Social Media Posts Draw Official Pushback

Earlier this week, the former engineer's series of posts warned that OpenAI and Anthropic were "barreling toward self-improving superintelligence, gambling with our lives." The thread predicted the imminent arrival of "superhuman systems" capable of infiltrating any network, upending any domain overnight, and seizing real power and resources.

The content exploded across social media in a matter of hours.

Michael directly referenced the viral thread during his remarks, attempting to cool the rhetoric. "It's easy to get into this doom loop — 'we're going to put 40% of Americans out of work, we're going to destroy humanity,'" he said. "It's just all your fears condensed into a well-written tweet."

The CTO expressed confidence that potential AI harms can be mitigated through free-market mechanisms, industry collaboration, and proactive engagement with government. Michael, a former senior executive at Uber Technologies Inc., is currently leading efforts to push widespread adoption of AI technology across U.S. military systems.

Procurement Dispute Turns Into Courtroom Battle

The friction between the Pentagon and Anthropic traces back to the company's insistence on additional safety requirements for AI tools intended for military use.

Michael led negotiations with the company, but talks collapsed earlier this year. The Defense Department rejected Anthropic's proposed terms, viewing them as an infringement on military authority, and subsequently designated the company as a national security threat to the U.S. supply chain.

The disagreement has since moved into federal courts. Anthropic filed lawsuits in both California and Washington, D.C., seeking to overturn the Pentagon's designation. A California court has so far ruled in favor of Anthropic, but the D.C. federal appeals court has yet to issue a decision.

Michael accused Anthropic of "judge shopping" by filing in the politically liberal Northern District of California. He characterized the D.C. Circuit case as "more significant" and "more national security-oriented," with a ruling expected in the coming months.

Workload Migration Nears Completion as Pentagon Sets Boundaries

On the operational front, the Defense Department has already moved roughly 90% of its classified AI workloads away from Anthropic, with the remaining portion expected to be fully transferred by the end of the month.

Speaking to reporters after the conference, Michael noted that "all work related to Maven Smart Systems or Palantir was migrated months ago," referring to the AI systems of the well-known defense technology company.

His comments stand in contrast to signals from Commerce Secretary Howard Lutnick, who had suggested that Anthropic had reconciled with the government after a "little kerfuffle."

Michael drew a clear line between the Defense Department's national security requirements and the perspectives of other federal agencies.

"I'm glad they've worked out their issues with other parts of government," he said, "but the Defense Department's position has always been clear: the way we partner with AI providers must serve warfighters and national defense. That means if we're going to use the software, we have to be able to use it for all lawful purposes without restrictions."