AI Competition Accelerates Across Military, Jobs, Web and Security

As artificial intelligence systems grow more powerful in 2026, their expanding footprint is exposing sharp tensions between technological capability and its real-world consequences. New data and reporting from multiple sectors paint a picture of an industry that extracts value at an unprecedented scale while returning relatively little, coinciding with structural shifts that are squeezing young workers, unsettling businesses, and accelerating geopolitical risks.

Cloudflare, which secures roughly one-fifth of the internet, has tracked how AI companies crawl the web compared with how often they refer users back to the original sites. The resulting “crawl-to-refer” ratio offers a stark ledger. Anthropic, the company that has built its brand on responsible development and constitutional AI principles, leads with a ratio of 8,800 to 1 in early April data. Its bots visit pages 8,800 times for every single referral sent back. OpenAI follows at 993 to 1. By comparison, Microsoft, Google, and DuckDuckGo operate with far more balanced figures. The pattern suggests that the grand bargain of the open web, in which creators make content available in exchange for traffic and visibility, is eroding. AI developers are effectively strip-mining the internet’s collective knowledge to train and improve their systems while diminishing the audience and revenue streams that sustain publishers, journalists, and independent sites.

This dynamic lands particularly hard on people entering the workforce. The underemployment rate for recent college graduates has climbed to 42.5 percent, the highest level since the depths of the pandemic. Gillian Frost, a 22-year-old quantitative economics major at Smith College, has applied to more than 90 jobs since September. She estimates that roughly 55 percent of applications receive automatic rejections and another 25 percent simply ghost her. “I feel helpless,” she told The Guardian. “No one seems to know how best to prepare due to the unique conflux of events occurring.” Frost and her peers are navigating a labor market strained by slower hiring, lingering post-pandemic adjustments, and employer expectations reshaped by generative AI tools that can draft reports, analyze data, and automate entry-level tasks once assigned to juniors.

Older graduates are not immune. Jeff Kubat, 31, returned to school in Minnesota for a master’s in accounting after eight years in accounts payable. He expected the degree to open doors. Instead, he has found the job search grueling even as he nears graduation. Across these stories runs a common thread: young people sense that the rules of career progression are being rewritten faster than institutions can explain or adapt to them. When AI systems can perform routine cognitive work, the premium on human judgment, creativity, and experience rises, but the pathways to acquire that experience are narrowing.

Enterprises are responding to a parallel set of concerns. A widening gap has opened between the frontier models offered by OpenAI and Anthropic and the needs of companies wary of handing sensitive customer data or intellectual property to outside APIs. Both firms maintain they do not train on enterprise data, yet repeated copyright lawsuits have left many executives uneasy. The result is growing interest in open-weights models from Google, Alibaba, Microsoft, and Nvidia. These releases, including Qwen 3.5, Gemma 4, and specialized MAI systems, are no longer viewed as research curiosities. Industry analysts describe them as credible enterprise platforms that allow organizations to run capable AI on their own infrastructure, preserving control and confidentiality. The shift suggests that even as the largest models grow more sophisticated, practical adoption may fragment toward smaller, specialized, or self-hosted systems.

Nowhere are the stakes higher than in national security. In September, China showcased autonomous drones capable of flying alongside fighter jets during a military parade attended by President Xi Jinping, Vladimir Putin, and Kim Jong-un. American officials viewed the display as evidence that the United States had fallen behind in unmanned combat systems. In response, the Pentagon has accelerated domestic programs. Anduril, the California defense startup, began production of its AI-backed Fury drone three months ahead of schedule at a new factory outside Columbus, Ohio. Russia and China are similarly investing in facilities to mass-produce advanced autonomous systems. Analysts have drawn comparisons to the early nuclear age, when rapid capability development outpaced efforts to manage escalation risks.

Taken together, these developments reveal a technology whose progress is outrunning the social, economic, and diplomatic frameworks needed to govern it. The same underlying advances in machine learning that let models ingest vast portions of the public web also enable weapons that make lethal decisions at machine speed. The same tools that reduce demand for entry-level analysis work can, in theory, increase productivity and create new categories of jobs, yet the transition appears prolonged and uneven. And the concentration of frontier capability in a handful of companies, each with distinct philosophies about safety and openness, leaves governments, news organizations, universities, and ordinary citizens negotiating with a narrow set of powerful actors.

Whether these tradeoffs can be mitigated through stronger data transparency rules, updated labor and education policies, or international agreements on military AI remains an open and urgent question. What is already clear is that artificial intelligence is not arriving as a neutral productivity layer. It is reorganizing incentives, power, and risk across the information economy, the job market, corporate technology strategy, and the international system simultaneously. The decisions made in the next few years about how to steer that reorganization will shape not only who benefits from AI but what kind of society it leaves behind.