Dispute Over AI Ethical Boundaries: Anthropic Risks $200 Million Pentagon Contract

Relations between the DoD and the startup Anthropic, considered one of the most ethical players in the AI market, have deteriorated sharply. The company led by Dario Amodei has opposed the use of the Claude model for purposes related to autonomous weapons and mass domestic surveillance. This assertiveness has met with an immediate response from the Pentagon, which is considering canceling a contract worth 200 million dollars. Secretary of Defense Pete Hegseth seeks unrestricted access to language models, motivated by the need to outpace China in the technological arms race. The context of the dispute is particularly significant because Anthropic's systems were allegedly already used in secret operations, including in Venezuela against the regime of Nicolás Maduro. However, the company's stance calls into question the commercial future of tech startups that try to combine profits from government contracts with rigorous security principles. The Pentagon is currently employing a hardball strategy, sending a clear signal to the entire Silicon Valley: either full cooperation or exclusion from lucrative public contracts. This situation creates a dilemma for competitors such as OpenAI or Google, who must balance the ethical expectations of their employees with geopolitical pressure. The conflict intensified after reports that the U.S. administration expects AI creators to abandon any protective barriers (so-called guardrails) in defense applications. Experts point out that Anthropic's unprecedented resistance could lead to a new dividing line in the industry, where some companies will focus solely on the civilian sector, losing access to billion-dollar federal funds. Since Project Maven in 2018, when Google employees protested against cooperation with the military on drone image analysis, relations between Big Tech and the Pentagon have remained tense and burdened by engineers' conscience issues.„We must integrate artificial intelligence faster and more effectively than our adversaries, which requires full access to the most advanced models without any questions.” — Pete Hegseth Positions in the AI dispute: Access to models: Limited by safety filters → Demand for full access (unrestricted); Main concerns: Ethics and existential risk → China's dominance in armaments; Value of cooperation: $200 million contract → Threat of complete contract termination