Report from the Wall Street Journal
In Brief – AI policy friction between the Trump Administration and AI giant Anthropic is emerging in a reported disagreement between the creator of the Claude chatbot and the Department of Defense. Claude is the leading AI chatbot for enterprise uses and the only top service certified for classified US government uses, resulting in its inclusion in major national security contracts as well as by smaller defense subcontractors. The Pentagon wants Anthropic to quickly agree that its tools can be used for all lawful use cases, a condition that AI rivals OpenAI, Google and xAI have agreed to in principle according to Emil Michael, the undersecretary of war for research and engineering, while Anthropic has not and objects to its technology being used for operations such as domestic surveillance and autonomous lethal activities. Pentagon leadership has cast “woke” tech companies as a liability, treats company restrictions as an impediment to military effectiveness, and is reviewing its partnership with Anthropic, which reportedly includes a threat to categorize the company as a supply-chain risk which would require national security contractors and vendors to certify that they don’t use its Claude models. An Anthropic spokesman said the company is having “productive conversations, in good faith,” with the Pentagon.
Context – No digital issue unites US conservatives more than the view that the digital giants repressed conservative viewpoints for years because they largely exist in a progressive, Bay Area monoculture. The potential for similar ideological bias within generative AI due to training data choices, directions given to human reviewers, and “guardrails” imposed by progressive company leaders quickly became a point of contention with AI systems. Along with pushing US AI policy firmly toward deregulation and investment and pitching US AI leadership as a national imperative to compete with China on economic, strategic and national security grounds, the second Trump Administration has made opposing “Woke AI” a core AI policy. The Administration’s AI leadership appears to see Anthropic as an adversarial policy leader.