A US appeals court has approved the Department of Defense blacklisting of Anthropic technology, ruling in a 2-1 decision that the Trump administration had authority to exclude the AI company from defense contracts even though Anthropic had no malicious intent. The case, decided by the US Court of Appeals for the District of Columbia Circuit, raises what the judges themselves called profoundly difficult questions about the appropriate relationship between AI safety and national defense.
The dispute centers on Anthropic refusal to enable certain Claude capabilities for military use. The company has maintained that its safety guidelines apply regardless of customer, a position that put it at odds with the Pentagon demand for unrestricted access to frontier AI models.
The majority opinion held that overly constrained AI models could cause military operations to fail and that the executive branch has broad discretion in determining which contractors serve national-security interests. The dissenting judge argued that the ruling effectively penalizes companies for maintaining safety standards.
The decision sends a chilling signal to the AI industry. Companies that build safety constraints into their models, a practice widely endorsed after a series of high-profile AI incidents, now face the prospect of losing government contracts if those constraints conflict with military requirements.
Anthropic said it is reviewing the decision. The ruling does not prevent the company from selling to commercial customers but effectively bars it from the federal market.
