On March 9, new developments occurred in the case of the US artificial intelligence company Anthropic suing the Department of Defense. More than 30 employees from OpenAI and Google DeepMind submitted a joint statement to the court, publicly supporting Anthropic’s position in the lawsuit against the US Department of Defense (DOD).
According to court documents, the statement stated that the government’s decision to list Anthropic as a “supply chain risk” was an inappropriate and arbitrary use of power and could have serious consequences for the entire artificial intelligence industry. Signatories include Jeff Dean, chief scientist of Google DeepMind and other industry heavyweight technical figures.
Previously, the U.S. Pentagon officially identified Anthropic as a "supply chain risk," a label that usually only applies to "foreign adversaries." According to reports, the trigger was Anthropic’s refusal to allow the Department of Defense to use its AI technology to conduct mass surveillance of U.S. citizens or for autonomous fire weapons systems; the Department of Defense insisted that as long as the use was “legitimate,” it should not be restricted by private contractors.
After Anthropic filed two lawsuits against the Department of Defense and other federal agencies, this "amicus brief" supporting its position immediately appeared in the case file, which shows the tension within the industry surrounding the boundaries of military AI applications. It is reported that the first person to disclose this opinion letter was "Wired" magazine.
In documents submitted to the court, these employees from Google and OpenAI emphasized that if the Pentagon is "no longer satisfied" with the terms of the existing contract with Anthropic, it can directly terminate the contract and switch to purchasing services from other leading AI companies, rather than "punishment" through supply chain risk identification.
In fact, at almost the same time that Anthropic was listed as a supply chain risk, the Department of Defense reached a new cooperation agreement with OpenAI, a move that also triggered strong protests from some employees within OpenAI. These employees are worried that the military's use of large commercial models to carry out sensitive testing will impact the industry's consensus on safety boundaries.
The opinion letter warned that if such "punitive" measures are continued, they will inevitably weaken the competitiveness of the United States in artificial intelligence and the broader scientific industry. The document further states that this approach will also have a chilling effect on public discussions about the risks and benefits of contemporary AI systems in the industry, and hinder the technology community from having candid debates on security issues.
In the current lack of a written public legal framework for the use of AI, the document believes that the various restrictions imposed by developers on the system through contractual terms and technical means are important safety valves to prevent catastrophic abuse of AI. The "red lines" Anthropic has drawn in its cooperation with the military, such as refusing to participate in mass surveillance and autonomous killing systems, are viewed by joint staff as reasonable and necessary security requirements rather than "an impediment to national security."
It is worth noting that many of the employees who signed the court statement have previously signed an open letter calling on the Department of Defense to revoke Anthropic’s supply chain risk identification, and asked their own company management to publicly support Anthropic and promise to refuse to provide their own AI systems to the military for unilateral use under undefended conditions.
Industry analysts believe that this case is not only about whether an AI company can adhere to its own safety bottom line at the contract level, but will also become a landmark event in how the United States draws a line between national security needs and AI security governance. As the lawsuit progresses, the game between technology companies, employees and the government over military AI applications is expected to further escalate.
Related articles:
Anthropic sues U.S. Department of Defense to block its inclusion on national security blacklist