Federal Agencies Bar Use of Chinese AI Model DeepSeek Over Data Security Concerns

The Biden administration announced a sweeping prohibition on the deployment of the artificial‑intelligence system DeepSeek across all federal departments, citing fears that the technology could funnel sensitive information to the Chinese government.

DeepSeek, a large‑language model developed by a Beijing‑based firm, has attracted attention for its advanced conversational abilities. However, officials warned that the model’s architecture allows for data transmission back to servers located in China, raising alarms about the potential exposure of classified or personally identifiable information held by U.S. agencies.

The ban follows a broader pattern of scrutiny directed at Chinese‑origin technology, including restrictions on certain facial‑recognition tools, cloud services, and previously on AI models from other firms. The administration’s decision reflects a growing consensus within the intelligence community that foreign‑controlled AI platforms pose a unique espionage risk, especially as they become more integrated into routine government workflows.

Under the new directive, agencies must cease all active use of DeepSeek and remove the model from any internal systems within a 30‑day window. The policy also bars the procurement of the software for future projects, and procurement officers are instructed to verify the provenance of any AI tools before acquisition.

Lawmakers have expressed bipartisan support for tighter controls on foreign AI, arguing that the rapid adoption of generative models outpaces existing security reviews. While some critics argue that the move could limit access to cutting‑edge technology, proponents contend that safeguarding national security outweighs the convenience of a single model.

Industry observers note that the ban may accelerate the development of domestically produced alternatives, as U.S. firms seek to fill the gap left by the exclusion of DeepSeek. The Department of Commerce is reportedly working with private sector partners to create vetted, home‑grown AI solutions that meet federal security standards.

The administration has not disclosed the specific evidence that prompted the action, but officials indicated that the decision was based on classified assessments of the model’s data‑handling practices. As the policy takes effect, agencies will likely conduct audits of their AI inventories to ensure compliance, while the broader debate over how to balance innovation with security continues to shape the federal technology landscape.