OpenAI and Google Employees Back Anthropic’s Legal Challenge Against the Pentagon
On Monday, Anthropic took legal action against the Department of Defense after being labeled a supply chain risk. Shortly after, around 40 employees from OpenAI and Google, including Jeff Dean, Google’s chief scientist and lead on the Gemini project, submitted an amicus brief supporting Anthropic. They expressed their concerns regarding the previous administration’s decision and the associated risks of the technology.
This development comes on the heels of a tumultuous period for Anthropic, which was designated a supply chain risk—typically a label for foreign entities considered a potential threat to national security—after the company maintained strict boundaries on military applications of its technology. These boundaries included a refusal to allow its AI for domestic mass surveillance and fully autonomous weapons. Following failed negotiations, tensions escalated, leading to public disputes and other AI firms stepping in to secure contracts for “any lawful use” of their technologies.
The supply chain risk designation not only bars Anthropic from military contracts but also puts other companies at risk if they utilize Anthropic products for Pentagon-related work, compelling them to abandon Claude to keep their lucrative contracts. However, Anthropic’s tools are already embedded in Pentagon operations; just hours after the Defense Secretary announced the designation, reports indicated that the U.S. military employed Claude in the operation that resulted in the death of Iran’s leader, Ayatollah Ali Khamenei.
The amicus brief argues that the supply chain risk designation is “improper retaliation that harms the public interest” and emphasizes that the concerns surrounding Anthropic’s red lines are significant and need addressing. It also suggests that these red lines merit reconsideration, highlighting that “mass domestic surveillance powered by AI poses profound risks to democratic governance — even in responsible hands” and that “fully autonomous lethal weapons systems present risks that must also be addressed.”
The authors of the amicus brief identified themselves as “engineers, researchers, scientists, and other professionals employed at U.S. frontier artificial intelligence laboratories.” They stated, “We build, train, and study the large-scale AI systems that serve a wide range of users and deployments, including in the consequential domains of national security, law enforcement, and military operations.” They clarified that they are not representing any single company but are sharing their insights as professionals with firsthand knowledge of the capabilities and risks of these systems.
Regarding domestic mass surveillance, the group noted that while data on American citizens is widely available through various sources like surveillance cameras, geolocation data, and social media, “what does not yet exist is the AI layer that transforms this sprawling, fragmented data landscape into a unified, real-time surveillance apparatus.” They explained that current data streams are isolated, but if AI were employed to integrate them, it could merge “face recognition data with location history, transaction records, social graphs, and behavioral patterns across hundreds of millions of people simultaneously.”
On the topic of lethal autonomous weapons, the group pointed out that these systems can be unreliable in unfamiliar conditions that differ from their training environments, meaning they “cannot be trusted to identify targets with perfect accuracy, and they are incapable of making the subtle contextual tradeoffs between achieving an objective and accounting for collateral effects that a human can.” They stressed the importance of human involvement in decision-making “before a lethal munition is launched at a human target,” especially since the reasoning behind these systems is often opaque to operators and even to their developers.
The authors concluded, “We are diverse in our politics and philosophies, but we are united in the conviction that today’s frontier AI systems present risks when deployed to enable domestic mass surveillance or the operation of autonomous lethal weapons systems without human oversight, and that those risks require some kind of guardrails, whether via technical safeguards or usage restrictions.”
