Anthropic Mythos AI Blocked: Pentagon Signs 7 Rival Deals

In a landmark AI policy decision, the Trump White House blocked a private company's AI from reaching 70 new customers in May 2026 — with no law passed and no court order signed. The target was Anthropic's Mythos, an advanced AI system that Anthropic's own researchers identified as capable of detecting and exploiting major cybersecurity vulnerabilities. That same period, the Pentagon signed AI contracts with seven of Anthropic's direct competitors — with no equivalent restrictions attached.

Behind this single policy contradiction sits a much larger story: the $1 trillion AI infrastructure buildout running at full speed, and a growing gap between the executives committing to it and the public that will live with the consequences.

AI Infrastructure: The $1 Trillion Bet No One Asked the Public About

Wall Street now projects $1 trillion in total AI infrastructure spending by 2027. BigTech capital expenditure (capex — the money companies spend on physical assets like data centers and specialized chips) was revised upward another 30% in 2026 alone to meet exploding compute demand. The semiconductor ETF called SOX (an exchange-traded fund that tracks chip makers including Nvidia, Broadcom, and TSMC) surged 50% in just the first four months of 2026.

Control over this infrastructure has consolidated into a remarkably small group:

• Nvidia & Broadcom — design the GPUs and ASICs (specialized processors that train and run AI models) powering virtually every major AI system
• TSMC — manufactures nearly all advanced AI chips in Taiwan
• Google TPUs — Google's custom-designed AI chips for internal workloads
• Microsoft, Amazon, Google, Meta — own the cloud infrastructure (the global data center networks) where AI runs at scale

These same companies hold equity stakes in venture-funded AI startups — meaning they profit from both the infrastructure those startups depend on and from the startups themselves. This is vertical integration (owning multiple layers of the same supply chain) at a scale regulators have not yet confronted. As the AI Supremacy newsletter puts it: "Just a few people at a few Big Tech companies control the future of AI, not anyone else."

The Mythos AI Policy Contradiction: Blocked for 70 Companies, Fine for Military Rivals

Anthropic planned to expand access to Mythos (a large language model — an AI system trained on massive datasets to perform complex reasoning, coding, and security analysis) to 70 additional organizations. The model's capabilities triggered internal safety reviews: Anthropic's researchers flagged its ability to detect and exploit real cybersecurity vulnerabilities.

The Trump administration blocked the expansion. No law authorized the move. No court issued an injunction. Legal observers describe this as the first known instance of a U.S. executive restricting commercial software distribution without any statutory or judicial basis — a precedent with no clear parallel in American technology policy.

In the same period, the Pentagon signed AI contracts with 7 competing vendors: SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, and AWS. None faced comparable restrictions. Anthropic's Mythos was deemed too risky for 70 civilian organizations while its competitors collectively collected government contracts without equivalent scrutiny.

The financial stakes are substantial. Anthropic is raising pre-IPO capital (funding rounds before a company lists on a public stock exchange) at a valuation floor of approximately $900 billion, with some projections crossing $1 trillion. Its annualized revenue run rate (ARR — projected yearly income based on recent monthly performance) sits at $44 billion as of mid-2026, per SemiAnalysis. A near-trillion-dollar company facing politically motivated distribution restrictions with no legal basis is a scenario venture capital risk models were not designed for.

AI Automation and the Labor Paradox: Manila Gained, San Francisco Lost

The AI jobs story turned out to be more specific — and more ironic — than the headline "AI destroys jobs" ever captured.

In the Philippines, 2 million call center workers remain employed as of 2026, up consistently since 2016 despite a decade of AI deployment in customer service. Economist Torsten Slok of Apollo explains the dynamic using the Jevons Paradox (the economic principle that making a resource cheaper tends to increase total consumption, not reduce it): "As AI makes call center work cheaper and faster, companies are buying more of it, not less. Lower cost per interaction does not mean fewer interactions."

U.S. programmers experienced the opposite dynamic. Programming jobs were growing at approximately 5% annually before ChatGPT launched in November 2022 — well above the overall labor market average. Since then, that growth rate has dropped by roughly 50%, confirmed by a March 2026 Federal Reserve study by economists Leland D. Crane and Paul E. Soto. The paper notes AI's overall labor impact remains "fairly ambivalent, uncertain and mixed," but the documented slowdown in one of the economy's historically strongest-growth professions is real.

One partial offset: Slok also notes that "sectors with the highest AI adoption rates have also seen the strongest growth in new business applications since 2022, showing that AI is lowering the barriers to starting a company." Whether AI-enabled solopreneurs (one-person businesses using AI to do work that previously required larger teams) can absorb displaced programmers at scale remains an open question — but it is the optimistic case.

Gen Z Is Not Buying the $1 Trillion AI Spending Narrative

Here is the number that should concern every executive approving a nine-figure AI budget: 22%. That is the share of Gen Z Americans who say they are genuinely excited about AI, per Gallup. Across the broader U.S. adult population, 50% say they are "more concerned than excited" about AI's role in society, per Pew Research.

That is a 78-percentage-point enthusiasm gap between the generation committing $1 trillion and the generation that will live with the consequences. The internet's mass rollout in the 1990s arrived with broadly positive public sentiment before the infrastructure was built. AI is achieving the opposite: deep public skepticism during the buildout, before most people have experienced its full effects.

The AI Supremacy newsletter, whose author has spent four years documenting AI geopolitics, frames the risk directly: "AI can destabilize a consumer-based economy, weaken democratic values and continue to warp Neo Capitalism." The author warns of a potential "AI populism" moment — a backlash from ordinary workers and consumers who see power concentrating rapidly while their own enthusiasm for the technology remains low.

If you want to cut through the trillion-dollar noise and find out what AI automation tools actually change in daily work, the AI for Automation guides cover practical applications across marketing, writing, and business — no trillion-dollar budget required. Watch which vendors the Pentagon keeps re-signing. Watch how the White House treats the ones it does not.

Related Content — Get Started with AI Automation | AI Automation Guides | More AI News