Pentagon Pressures Anthropic on AI
The Pentagon's Chief Technology Officer has publicly urged AI company Anthropic to “cross the Rubicon” and fully support military use cases for its technology. The statement comes amid an ethics dispute where Anthropic has shown reluctance to provide its Claude large language model for autonomous weapons or mass surveillance, in contrast to competitors like OpenAI and Google who have complied with Department of Defense requests. The Pentagon's stance suggests it will prioritize mission utility over vendor's ethical concerns.
- Anthropic’s public ethical stance is governed by its "Responsible Scaling Policy," which uses a system of AI Safety Levels (ASLs) to manage risks as models become more capable and includes a public commitment not to deploy models that could cause catastrophic harm without adequate safeguards. - The dispute is not a blanket refusal of all defense work; Anthropic is reportedly comfortable with its Claude model being used for intelligence analysis or logistics, but draws the line at applications for lethal autonomous weapons and mass domestic surveillance. - In the summer of 2025, the Pentagon awarded contracts for "frontier AI projects" worth up to $200 million each to Anthropic, OpenAI, Google, and xAI; Anthropic's contract is now reportedly under review. - The Department of Defense has a formal Responsible Artificial Intelligence (RAI) Strategy and Implementation Pathway, which guides the ethical development and deployment of AI across the department. To operationalize this, the Chief Digital and AI Office (CDAO) developed the RAI Toolkit, a resource for developers to ensure their capabilities align with the DoD's ethical principles. - The Pentagon has launched a platform called GenAI.mil to provide AI tools to its roughly 3 million personnel; the platform launched with Google's Gemini models and has plans to integrate offerings from OpenAI and xAI. - This is not the first time a major tech company has clashed with the Pentagon over AI ethics; Google, an early contractor on the military's Project Maven AI-powered image analysis program, withdrew in 2018 following employee protests. - The Pentagon could designate Anthropic a "supply chain risk" if the dispute is not resolved, a label typically reserved for foreign adversaries that could impact Anthropic's ability to work with other government contractors. - The Pentagon's Undersecretary of Defense for Research and Engineering, Heidi Shyu, serves as the department's Chief Technology Officer and has publicly identified "Trusted AI and Autonomy" as one of the critical technology areas for maintaining U.S. military advantage.
