OpenAI secures classified AI deal as Washington sidelines Anthropic

OpenAI chief executive Sam Altman confirmed that the company has reached a strategic agreement with the U. S. Department of War to deploy its artificial intelligence models on the ministry’s classified network, signalling a major shift in the technology sector’s engagement with national security infrastructure and government contracts. The contract embeds safety principles that bar mass domestic surveillance and fully autonomous weapon-control, reflecting a delicate balance between innovation and ethical restraint as the administration redefines its approach to AI governance.

The new arrangement follows a dramatic policy move by President Donald Trump, who ordered all federal agencies to cease using AI services from Anthropic, a leading rival, after the company declined to accept the Pentagon’s terms that would grant unrestricted access to its Claude models for lawful purposes. Defence Secretary Pete Hegseth has designated Anthropic a national security supply-chain risk, effectively terminating its $200 million contract and barring government contractors from engaging with the firm. Anthropic has vowed to legally challenge the designation as “legally unsound” and unprecedented.

Altman’s announcement underscores OpenAI’s expanding footprint in national security applications, even as he called for industry-wide norms. He said that the Department of War “showed a deep respect for safety” and that the company and the ministry have agreed on guardrails that uphold human oversight and prohibit domestic surveillance and autonomous weapons systems. OpenAI will deploy its models through cloud networks with field engineers to monitor usage and ensure compliance with these principles. The firm has urged the Department of War to extend similar contractual terms to other AI providers, an apparent appeal to de-escalate tensions within the broader artificial intelligence ecosystem.

The backdrop to these developments is an intensifying clash over how cutting-edge AI can be used in defence and national security. Anthropic, founded by former OpenAI leaders and known for its emphasis on safety and constitutional AI, had positioned Claude as compliant with restrictions on its deployment in surveillance and lethal autonomous systems. This position, however, collided with the Pentagon’s demand for unrestricted use in any lawful military application, prompting a standoff that erupted into a wholesale ban on Anthropic’s technology across federal agencies.

Industry responses to the federal directive have been mixed. Some leading figures in the tech sector rallied to Anthropic’s defence, voicing concern that government coercion undermines private sector autonomy and ethical safeguards. Others have portrayed the administration’s actions as essential to ensuring that AI capabilities are fully available to defence operations. The dispute has exposed deeper fault lines about the intersection of technology, ethics and national security, with major AI labs like Google and subsidiaries of Elon Musk’s xAI also navigating their own Pentagon approvals or partnerships.

Amid these policy shifts, OpenAI is also advancing its commercial and strategic alliances. The company has agreed a strategic partnership with Amazon that could involve up to $50 billion in investment, with significant initial funding and commitments contingent on meeting milestones. This collaboration is expected to bolster OpenAI’s infrastructure through expanded access to cloud computing power and drive innovation across generative AI offerings, shaping competitive dynamics against rivals such as Microsoft and Google.

Washington’s broad directive to halt use of Anthropic’s technology represents a rare assertive stance by the federal government in dictating the terms on which private AI firms can participate in national security projects. Officials have justified the move on grounds of ensuring unfettered access for defence purposes, even as it raises questions about precedent and industry self-regulation. Anthropic’s legal challenge and the broader industry debate may force policymakers to re-evaluate how ethical constraints and safety priorities coexist with security imperatives.