A convergence of forces—Pentagon ultimatums, corporate policy rollbacks, and market panic—suggests the AI industry's moment of truth has arrived.
The Pentagon's Friday Deadline
On Tuesday, February 25, 2026, Defense Secretary Pete Hegseth met with Anthropic CEO Dario Amodei at the Pentagon for what was described as a "tense" face-to-face conversation. The outcome: a deadline. By Friday at 5:01 PM, Anthropic must agree to allow its Claude AI models to be used for what the Pentagon is calling "unrestricted military applications"—specifically, domestic surveillance and fully autonomous weapons systems.
This is not a minor negotiation. The Pentagon has made clear it will invoke the Defense Production Act to force compliance if necessary, and has threatened to declare Anthropic a "supply chain risk"—a designation that would effectively blacklist the company from federal contracting. At stake is a $200 million contract, and more importantly, the precedent for how AI companies interact with military and intelligence agencies.
Anthropic's position has been notably defiant. The company has stated that its safety guidelines—which prohibit the use of Claude for mass surveillance or for weapons that can identify and engage targets without human oversight—are non-negotiable.
The Quiet Retreat from Safety Pledges
While the Pentagon was delivering its ultimatum, Anthropic was executing a different kind of rollback—one that received far less attention but may prove more consequential in the long run.
On February 25, 2026, the same day as the Pentagon meeting, Anthropic released an updated version of its Responsible Scaling Policy. The changes were significant. The company removed its previous commitment to halt the development of AI models if they outpaced the company's ability to ensure their safety.
In its place, Anthropic now offers what it describes as "nonbinding but publicly-declared" goals—objectives the company will "grade itself on," but which carry no contractual or legal weight.
The Citrini Panic
On Monday, February 24, U.S. stocks slid sharply after a research memo from Citrini Research went viral across trading desks. The memo, titled "The 2028 Global Intelligence Crisis," was explicitly framed as a scenario, not a prediction.
The memo's core thesis: if AI automates high-earning knowledge workers—software engineers, financial analysts, lawyers—at the rate that current trajectory suggests is possible, the consumption engine that drives the economy sputters. Mortgage defaults rise. Private credit collapses.
Citrini's scenario predicted unemployment could reach 10% by mid-2028, with the S&P 500 falling 38% from its October 2026 highs.
The LinkedIn Reality Check
Against these dramatic headlines, the labor market data from LinkedIn offers a more nuanced—and somewhat contradictory—picture.
According to LinkedIn's January 2026 Labor Market Report, AI is not destroying jobs on net. The platform's data shows that over the past three years, mentions of AI in job listings have increased by more than 600% in the United States. More concretely, LinkedIn identifies 1.3 million new "AI-enabled jobs" globally—roles like AI engineers, forward-deployed engineers, and data annotators that did not exist at scale five years ago. In the data center sector alone, there were over 600,000 new AI-enabled positions created in the past year.
This is not a story of mass displacement—at least not yet. It is a story of rotation. The labor market is shifting, not shrinking. New collar roles are emerging that blend knowledge work, advanced technical skills, and distinctly human capabilities. According to the U.S. Bureau of Labor Statistics, 60% of new jobs by 2030 will come from occupations that do not typically require a degree. The LinkedIn report also notes that companies can grow their AI talent pipeline by as much as 8.2x globally if they focus on skills over formal degrees—a staggering multiplier that suggests the talent constraint is artificial, not structural.
The data suggests a profound reallocation rather than elimination. Companies are not simply cutting headcount; they are restructuring entire workflows around AI capabilities. The accountant becomes an AI-augmented analyst. The coder becomes a prompt engineer. The designer becomes an AI collaboration specialist. Each role transforms, but the human remains essential—not as a cost center to be eliminated, but as an adaptor to be retained.
And yet, the underlying anxiety is real. LinkedIn's own research finds that 56% of professionals plan to job hunt in 2026, while 76% say they don't feel prepared to find a new job in the current market. The disconnect between aggregate job creation and individual insecurity is striking. Even as LinkedIn counts 1.3 million new AI-enabled roles, workers sense something is shifting beneath their feet.
The Broader Implications
The convergence of these events—the Pentagon's hardball tactics, Anthropic's policy retreat, the market's sudden jitters, and the labor market's quiet transformation—paints a picture far more complex than the usual narratives allow. We are not simply witnessing the growth of AI, nor its decline, nor even its regulation. We are witnessing the emergence of AI as a political, economic, and social force in ways that transcend any single framework.
The Pentagon's demands represent the militarization of AI—government's attempt to harness powerful technology for national security purposes, with all the ethical compromises that entails. Anthropic's policy shift represents the commercialization of ethics—the willingness to abandon stated principles when they become inconvenient. The Citrini memo represents the financialization of anxiety—the translation of diffuse fears into concrete market action. And the LinkedIn data represents the humanization of disruption—the recognition that real people, with real skills and real livelihoods, are navigating a transformation that no one fully understands.
These are not separate stories. They are facets of a single phenomenon: AI is no longer just a technology. It is a fault line running through every institution that relies on human labor, human judgment, and human trust. The question is not whether AI will change the world—the evidence suggests it already has. The question is whether that change will be managed, equitable, and humane, or whether it will be chaotic, concentrated, and cruel.
What Comes Next
These three threads—the Pentagon's demands, Anthropic's policy retreat, and the Citrini panic—do not tell a single coherent story. They are, in some ways, contradictory. If AI is creating 1.3 million new jobs, why should we fear a 2028 unemployment crisis? If Anthropic is truly committed to safety, why is it quietly weakening its safety commitments? If the Pentagon wants AI so badly, why is it threatening to blacklist the only company that has maintained meaningful ethical guardrails?
The answer is that we are not dealing with a single AI industry, but with multiple AI industries with competing interests and incompatible worldviews. There is the AI that governments want: powerful, compliant, unrestricted, and useful for surveillance and warfare. There is the AI that companies want to sell: transformative, profitable, and accompanied by enough safety theater to ward off regulatory scrutiny. There is the AI that markets have priced in: endlessly productive, endlessly growing, endlessly scalable. And there is the AI that workers fear: a replacement for their skills, their livelihoods, their relevance.
What we are witnessing is not the resolution of these tensions, but their acceleration. The next few months will determine which AI industry wins out—or whether, as the Citrini scenario suggests, they all collapse together. The Pentagon will get its answer from Anthropic by Friday, February 27, 2026. Markets will continue to reprice AI risk. Workers will continue to adapt, or not, to the new reality. And the rest of us will watch, somewhere between fascination and dread, as the story unfolds.
One thing is clear: the era of AI companies being able to have it all—profit and safety, government contracts and ethical commitments, market optimism and worker security—is ending. The question is what replaces it, and whether anyone will be in control when it does. The next chapter of the AI story will not be written by algorithms alone. It will be written by the choices we make—individually, institutionally, and collectively—in the weeks and months ahead.
Word count: approximately 2,113 words