Pentagon Blacklist Failed to Block China’s Biggest Tech Firms From Accessing U.S. AI

OpenAI and Google supplied advanced artificial intelligence services to Singapore-based subsidiaries linked to Alibaba, Baidu and Tencent despite the Chinese companies appearing in the U.S. Department of Defense blacklist, according to statements the companies provided to the Financial Times.

The transactions were legal under current U.S. export regulations because the Pentagon’s Chinese Military Companies list does not automatically prohibit American companies from selling AI services. Instead, the designation identifies firms the Defense Department says have ties to China’s military, while separate export-control rules determine which technologies may be sold.

The disclosures expose an increasingly important distinction as Washington expands efforts to limit China’s access to advanced technology. While the United States has tightened restrictions on AI chips and semiconductor equipment, commercial access to frontier AI models remains governed by a different set of rules.

OpenAI and Google Say the Transactions Complied With U.S. Rules

OpenAI said it provided access in accordance with U.S. law and its own usage policies. Google likewise said it complies with applicable export controls and sanctions requirements when offering AI services.

Both companies said the customers were Singapore-based subsidiaries rather than entities operating directly in mainland China.

Neither company suggested the Defense Department’s designation, by itself, prevented those organizations from accessing their AI models.

The companies’ responses underscore that appearing on the Pentagon’s blacklist is not the same as being subject to a comprehensive export ban.

Anthropic Chose a More Restrictive Policy

Anthropic has taken a different approach.

The company told the Financial Times it does not provide its most advanced AI models to organizations based in China, even where doing so may be permitted under U.S. law.

That policy stands in contrast to OpenAI and Google, whose commercial decisions follow existing export regulations rather than imposing broader company restrictions.

The difference illustrates how AI developers are adopting their own risk policies as governments continue to refine rules governing frontier AI.

Why Singapore Became Part of the Story

Singapore has become a major regional hub for multinational technology companies serving customers across Asia.

According to the companies, the AI services referenced in the report were delivered through Singapore-incorporated subsidiaries connected to Alibaba, Baidu and Tencent rather than directly to mainland Chinese entities.

The structure reflects a common corporate arrangement for international businesses operating across multiple jurisdictions. It also highlights how AI services, unlike physical semiconductor shipments, can raise different regulatory questions under existing U.S. export controls.

The Policy Debate Is Likely to Continue

The disclosures arrive as U.S. policymakers continue debating how to regulate access to frontier AI technologies.

Washington has imposed increasingly stringent restrictions on advanced semiconductor exports to China, but cloud-based AI services remain subject to a different regulatory framework.

The episode has drawn attention to the distinction between military-related designations issued by the Defense Department and export restrictions administered by the Commerce Department. Whether those rules should be aligned more closely is likely to remain part of the broader debate over U.S. AI policy and technology competition with China.