The Pentagon announced Friday that it has secured agreements with seven technology corporations to integrate their artificial intelligence into its classified computer networks, enabling the military to deploy AI-powered capabilities in its warfighting operations. This move comes as public resentment toward AI is reportedly "boiling over" in America, with only 30 percent of citizens trusting their government to regulate the rapidly scaling technology.
The Washington Post characterized the Pentagon's agreements as part of a fundamental shift in U.S. AI policy. Dean Ball, a former senior adviser on AI policy for President Donald Trump’s administration, echoed this assessment in a separate Washington Post newsletter, stating that a fundamental shift in AI policy is underway.
The State's War Machine
The U.S. industrial strategy is redirecting tech investments from established coastal centers to new inland sites in Arizona and Texas. This shift is fostering a burgeoning AI ecosystem in the Sun Belt, as detailed in a new report by Eva Dou.
Politico reported on the newly released Stanford AI Index Report, now in its ninth year, which tracks AI’s speed, scale, and influence. The report found that the world is adopting artificial intelligence at a pace so rapid that its capabilities cannot be measured, its progress assessed, or its training models understood.
The 2026 Stanford report revealed that generative AI tools are being adopted faster than the internet. This rapid integration occurs amidst widespread public distrust, with the 30 percent of Americans trusting government regulation being the lowest rating among all 30 countries surveyed.
Sha Sajadieh, Stanford’s AI Index lead, noted that where there are high levels of adoption and enthusiasm for AI, there also tends to be a high level of trust that governments will protect and regulate the technology effectively. Sajadieh contrasted this with the U.S., where there is not as much enthusiasm or adoption, and significantly less trust in the government to regulate AI in a way that might protect the public.
Sajadieh offered a personal hypothesis, stating that institutions were not built for technological transformation that happens at such speed. He added that the adoption of generative AI tools has outpaced the internet and the personal computer, moving faster than anything before, and that education systems and governance policies are not designed to keep up.
Capital's New Frontiers
Regarding the aggressive development and investment in AI by America despite public skepticism, Sajadieh suggested that coverage of the technology often falls into two camps: "hype headlines" about new models' capabilities and headlines concerning "mass displacement" and disruption. He stated that a lack of objective information for the U.S. public leads to a void, causing people to turn to easily digestible but potentially unreliable sources.
Sajadieh emphasized that policymakers need better data to support regulation, which could help restore public confidence. He specified that transparency from "frontier labs" is paramount, as parameters for how models are developed and what they are trained on are often not disclosed, making them increasingly "black box" systems.
In a comparison of global AI output, Sajadieh noted that the U.S. and China lead in different ways. The U.S. leads in private investment, pouring more dollars into AI companies than any other nation, which is reflected in the output of notable models. China, conversely, has a different investment structure, with significant government funding or public-private investment.
Despite these differing investment models, the capabilities of the AI models produced by both countries are converging. Sajadieh also pointed out that China has a more open AI ecosystem than the U.S., suggesting it could be a model for development. Other nations, such as South Korea, Singapore, and Switzerland, are also developing highly capable models and maintaining talent pipelines, with South Korea focusing on its own AI sovereignty through both English and Korean language models.