Pentagon Chief to Meet with Anthropic CEO Regarding Access to AI Claude

U.S. Secretary of Defense Pete Hegseth has called for a direct meeting with Anthropic CEO Dario Amodei. The talks will take place on Tuesday at the Pentagon headquarters and aim to break a prolonged deadlock in contract negotiations. The dispute concerns the conditions under which the Department of Defense could use the advanced artificial intelligence model Claude, developed by Anthropic. Anthropic was founded in 2021 by Dario Amodei and a group of researchers who left OpenAI. From the start, it declared an ethical approach to AI development, with a strong emphasis on safety and built-in limitations. Its flagship model, Claude, competes with products from giants like OpenAI's ChatGPT and Google's Gemini. The main point of contention are the so-called guardrails, the built-in technological safeguards and limitations that Anthropic has imposed on its technology. The company refuses to remove them entirely for military use, which has met with sharp opposition from the Pentagon. In response, the department is threatening to designate Anthropic as a supply chain risk. This is a legal and procedural category typically applied to firms from countries considered adversaries or allies of U.S. adversaries. Such a threat underscores the seriousness of the conflict and the military's determination to gain full access to the technology. According to reports from Axios, Claude is the only available AI model in secret, classified military systems and is seen as the most capable for performing sensitive defense and intelligence tasks. The Pentagon does not want to lose access to this tool but is simultaneously furious at the company's resistance. „Claude is the only AI model available in the military's classified systems, and the most capable model for sensitive defense and intelligence work.” — Axios In the background of the case is also an alliance announced in November 2024 between Anthropic and the company Palantir Technologies. The partnership was meant to enable intelligence and defense agencies to 'operationalize the use of Claude on Palantir's AI platform.' The current dispute could threaten the realization of this cooperation or change its scope. The meeting between Amodei and Hegseth is a last-ditch effort to find a compromise. Sources indicate that the parties are entering it 'from completely different positions.' The military is pushing for full freedom, while the company maintains that certain ethical and safety limitations are an inseparable part of its product. The outcome of these negotiations will have significance not only for these two entities but also for the broader framework of cooperation between the defense sector and the commercial AI industry, which balances innovation with responsibility.