Pentagon Official Dismisses 'AI Will Wipe Out Humanity' Warning After 150 Million Views

Days after a former engineer at Anthropic publicly warned that industry insiders believe the technology 'could kill everyone alive before this generation ends,' a high-ranking U.S. Department of Defense official has pushed back against rising concerns about AI risks.

Emil Michael, the Pentagon's Chief Technology Officer and top official overseeing research and engineering, addressed the social media post by Jacob Coxon, a former Anthropic engineer, during a defense industry conference in Washington on Thursday (the 10th).

Coxon's post has now been viewed more than 150 million times, encapsulating widespread anxiety about uncontrollable AI systems dominating a 'dystopian' world.

Michael said: 'It's easy for people to fall into this apocalyptic cycle—thinking, “We're going to put 40% of Americans out of work, we're going to destroy humanity.” But really, it's fear of data centers, fear of change—all condensed into one well-written tweet.'

Earlier this week, Coxon posted a series of warnings, claiming OpenAI and Anthropic are racing toward self-improving superintelligence, 'betting with human lives,' and predicting that 'systems surpassing humans will soon emerge, capable of hacking any target, overturning any domain overnight, and seizing real power and resources.'

Michael, formerly a senior executive at Uber Technologies, now leads the Pentagon’s efforts to accelerate AI adoption across the U.S. military. He argued that the potential harms highlighted by Coxon can be mitigated through free markets, industry collaboration, and coordination with government.

The Pentagon’s push for AI integration has sparked a fierce dispute with Anthropic. The core disagreement lies in Anthropic’s demand for additional safety safeguards when its AI tools are used by the U.S. military.

Michael previously helped lead negotiations with Anthropic, but talks broke down earlier this year. The Pentagon rejected Anthropic’s conditions, viewing the restrictions as an infringement on military authority, and designated the company a supply chain risk for the United States.

The dispute has since moved to federal court. Anthropic filed lawsuits in both California and Washington, D.C., seeking to overturn the Pentagon’s designation. A judge in California ruled in Anthropic’s favor, but the U.S. Court of Appeals for the D.C. Circuit has yet to issue a ruling.

Michael called the D.C. Circuit case 'more important' and insisted that, without a preliminary injunction, Anthropic remains a supply chain risk. Anthropic did not immediately respond to requests for comment on Michael’s remarks on Thursday.

Michael accused Anthropic of 'forum shopping'—deliberately choosing the Northern District of California, a court perceived as politically liberal. He argued that the D.C. Circuit case carries greater significance and expects a decision in the coming months from what he described as a jurisdiction 'more focused on national security.'

Later, speaking to reporters, Michael said the Pentagon has already migrated approximately 90% of its classified AI workloads away from Anthropic and expects to complete the remaining transition by the end of this month.

He stated: 'All Maven Smart Systems or Palantir-related work was moved out several months ago.' He referred to the AI systems developed by Palantir, a prominent defense technology firm.

Despite U.S. Commerce Secretary Howard Lutnick stating that Anthropic has reconciled with the government after a 'small spat' with the Trump administration, Michael emphasized that the Pentagon’s national security requirements differ from those of other federal agencies.

'I’m glad they’ve resolved their issues with other parts of the government,' Michael said. 'But we’ve been very consistent in our position—that we must partner with AI vendors in a way that benefits warfighters and national defense.'

He added: 'That means if we decide to use the software, we must be able to use it without restriction for all lawful purposes.'