US Orders Removal of DeepSeek From Federal Civil Agency Systems 2026

The US government has ordered the removal of DeepSeek AI from federal civil agency systems, marking a significant escalation in the ongoing tech rivalry between the United States and China. The directive, dated July 24, 2026, and published in the Federal Register, requires executive branch civil agencies to cease new acquisitions of DeepSeek products and services and to eliminate existing instances from their computer systems.

This move comes after months of scrutiny surrounding DeepSeek, the Chinese company that burst onto the scene with its R1 model, challenging perceptions of which nation holds the most capable or cost-effective AI technology.

What Exactly Has the US Ordered?

The order applies specifically to civil agencies of the federal executive branch. These entities must halt any new purchases of DeepSeek-developed products and services and remove any such technology already present in their IT infrastructure. The measure allows for exceptions upon request and explicitly excludes national security systems and compartmentalized information systems.

The scope extends beyond the consumer-facing chatbot. DeepSeek's models can be accessed via its API and integrated into third-party applications and services, meaning its technology could be embedded in tools under different brand names. This complicates detection and removal efforts for agencies.

Why This Matters: The Broader Tech Rivalry

DeepSeek has been at the center of a dispute that goes far beyond technical benchmarks or pricing. Since the release of R1, the company has faced increasing scrutiny from governments, regulators, and US businesses amid heightened technological competition.

The US order reflects growing concerns over data security, privacy, and potential foreign influence in critical government systems. It also signals a willingness to take concrete action against AI models originating from countries perceived as strategic competitors.

For federal agencies, the directive means auditing their software supply chains to identify any DeepSeek dependencies, including those buried in third-party tools. This could disrupt workflows and require significant resources to ensure compliance.

Impact on Businesses and Users

While the order directly targets federal civil agencies, its ripple effects may reach the private sector. Companies that contract with the government may need to reassess their use of DeepSeek to avoid conflicts. Additionally, the directive could influence corporate policies regarding AI tools from foreign adversaries.

For individual users, the order does not restrict personal use of DeepSeek. However, it underscores the importance of understanding the origins and data handling practices of AI services. Enterprises should evaluate whether their AI vendors have ties to restricted entities.

• Federal civil agencies must remove DeepSeek from their systems and stop new acquisitions.
• National security systems are exempt from the order.
• Third-party products embedding DeepSeek via API are also covered.
• Exceptions can be requested, but the default is removal.
• Private sector may face indirect pressure to follow suit.

Comparison: DeepSeek vs. Other AI Models in Government Use

As the table shows, the US order singles out DeepSeek based on its Chinese origin, while domestic models face standard regulatory scrutiny but are not banned outright.

What to Watch Next

The implementation of this order will be closely watched. Agencies must now inventory their systems, identify DeepSeek usage, and plan for removal. The exception process may reveal which use cases are deemed critical enough to warrant continued access. The background to this story is set out in Germany Approves Spy Agency Reform in Security Revolution. A related thread runs through Mystery AI Model Sparks Developer Frenzy – DeepSeek’s Secret?. Earlier coverage in Taiwan AI-Assisted Cyber-Attack Hits Government Agencies shows how the argument developed.

Furthermore, the order could prompt other countries to take similar actions, potentially fragmenting the global AI landscape. Companies developing AI tools should prepare for a world where geopolitical considerations increasingly dictate technology adoption. A closer look at the same subject appears in US crises deepen as Trump administration cuts vital agencies. Anyone following the wider picture can start with Comer Insurance Agency: Services, Coverage & Reviews. The same theme surfaces in Canada Wildfire Crisis Demands National Agency Now, which takes a different angle.

For now, the message is clear: the US government is drawing a line in the sand on AI from strategic competitors. Context on how this played out before sits in AI Safety Regulation: Can Rules Stop AI Systems Killing Us. A companion piece, Why OpenAI and Anthropic Support AI Regulation in Australia, approaches the same subject from another angle.