Pentagon’s Futurist BOLTS to Blacklisted AI Firm

Close-up of a smartphone displaying the word 'ANTHROPIC' with a blurred portrait in the background

The Pentagon’s most secretive strategic mind just walked out the door and straight into the boardroom of an AI company locked in a bitter feud with his former employer.

Story Snapshot

James Baker, who ran the Pentagon’s Office of Net Assessment for a decade, joined Anthropic as strategist-in-residence
Baker warns AI presents a “civilizational” threat to U.S. institutions requiring urgent adaptation in a “tight time window”
The hire comes during Anthropic’s forced six-month federal withdrawal and ongoing Pentagon ultimatum over AI model access
Baker will focus on analyzing AI’s impact on U.S.-China competition while the company resists unrestricted government use of its technology

The Pentagon’s Oracle Switches Sides

James Baker spent ten years running the Office of Net Assessment, the Pentagon’s most enigmatic strategic planning operation. From 2015 to 2025, he advised defense secretaries and national security advisors on how emerging technologies would reshape American military power. The ONA, established in 1973, earned its reputation as the Pentagon’s intellectual engine by focusing on threats most officials would not confront for decades. Baker’s departure was not voluntary. The Trump administration shuttered the office in 2025, ending a half-century tradition of long-range strategic forecasting just as artificial intelligence began reordering the global balance of power.

Anthropic announced Baker’s hiring in early May 2026, positioning him to lead analysis of AI’s effects on American institutions and the escalating technology race with China. The timing raises eyebrows. Anthropic currently executes a presidential order forcing its withdrawal from federal service for six months. Defense Secretary Pete Hegseth issued an ultimatum to CEO Dario Amodei in February 2026, demanding unrestricted access to Anthropic’s AI models for military applications or face severed ties. The Pentagon designated Anthropic a supply chain risk earlier this year, casting the company as an obstacle to national security rather than an asset.

When Strategy Becomes Survival

Baker’s public statements since joining Anthropic reveal a man consumed by urgency. He describes AI as a multi-decade structural problem that threatens the long-term viability of American institutions during both war and peace. The “civilizational” language distinguishes his perspective from typical Silicon Valley hyperbole or Pentagon threat inflation. Baker spent a career identifying gaps between technological change and institutional adaptation. His warning carries weight precisely because his job was finding threats others missed. The United States has a tight window to adjust, he argues, before AI fundamentally alters which institutions can function and which collapse.

This framing should worry anyone paying attention. Baker is not discussing battlefield advantages or economic productivity gains. He is questioning whether the basic structures of American government, military command, and civil society can survive contact with artificial intelligence without radical transformation. The Pentagon bet on incremental AI integration for targeting systems, drone coordination, and cyber operations. Baker’s emphasis on institutional viability suggests that approach misses the scale of disruption ahead. If the man who advised four defense secretaries on existential threats now works for a private AI lab, the center of gravity in national security strategy has shifted outside government.

The Ethics Versus Security Collision

Anthropic built its brand on responsible AI development, positioning itself against competitors willing to deploy models without extensive safety testing. That stance now collides with Pentagon demands for unrestricted access to cutting-edge AI for military applications. The February ultimatum from Hegseth epitomizes the clash: national security requires technological superiority, and the Defense Department will not tolerate a domestic company withholding capabilities that could determine whether America wins or loses a war with China. Anthropic’s resistance signals either principled concern about autonomous weapons and domestic surveillance or a catastrophic misunderstanding of the stakes.

Baker’s hire suggests Anthropic seeks credibility on national security questions while maintaining ethical boundaries. Adding the former ONA director provides strategic heft to arguments that responsible AI development serves long-term American interests better than rushing unproven systems into warfare. Yet the hire also highlights a troubling pattern. Top government experts on AI and national security increasingly migrate to private companies offering higher pay and fewer bureaucratic constraints. The Pentagon loses institutional knowledge and strategic vision to firms that may prioritize shareholder value over national defense when conflicts arise. Baker’s move accelerates the privatization of strategic thinking at precisely the moment government needs it most.

Former head of 'Pentagon's think tank' joins Anthropic – Defense One https://t.co/wpFq9x1hnR

— Jim Kaskade (he/him) (@jimkaskade) May 1, 2026

The standoff between Anthropic and the Pentagon exposes deeper questions about who controls transformative technology in a democratic society. The Defense Department argues correctly that AI will determine military outcomes in any future great power conflict. Surrendering leadership to China because American companies refuse cooperation would constitute strategic suicide. Anthropic counters that unrestricted military AI risks catastrophic accidents, erosion of civil liberties through domestic surveillance, and autonomous weapons that remove human judgment from life-and-death decisions. Both positions contain truth, which is why the conflict defies easy resolution and why Baker’s expertise matters.

The China Factor Nobody Wants to Discuss

Baker’s role explicitly includes analyzing U.S.-China AI competition, a contest that will define the 21st century as much as nuclear weapons shaped the 20th. China does not handicap its AI companies with ethical debates about autonomous weapons or surveillance. The Chinese Communist Party treats AI as an instrument of state power, directing massive resources toward military applications while harvesting data from a surveilled population of 1.4 billion people. American hand-wringing about responsible AI development looks quaint against an adversary that views such concerns as weakness to exploit.

This reality makes Anthropic’s resistance to Pentagon demands harder to defend from a national security perspective, even if the ethical concerns have merit. Baker’s civilizational framing may bridge the gap by redefining success. If AI truly threatens institutional viability, then reckless deployment to match Chinese speed could destroy what America aims to protect. Maintaining institutions capable of democratic governance and individual liberty while competing with an authoritarian surveillance state requires different strategies than a pure technology arms race. Whether Baker can translate that insight into policy recommendations that satisfy both Anthropic’s ethics and Pentagon’s security requirements remains uncertain.