WASHINGTON — The US Cybersecurity and Infrastructure Security Agency widened its ban on DeepSeek on Wednesday, urging operators of critical infrastructure to remove the Chinese artificial-intelligence model from networked systems.

The move followed Monday's order requiring civilian federal agencies to purge DeepSeek from government IT systems within a set deadline. CISA cited concerns that the model's data could be routed to servers in China and exploited for espionage or supply-chain compromise.

The directive named the energy, water and telecommunications sectors as priorities, according to a CISA advisory published alongside the Department of Homeland Security. The guidance recommended that operators inventory any DeepSeek-derived applications and block associated network traffic. It stopped short of a binding mandate for private firms.

DeepSeek, developed by the Hangzhou-based startup of the same name, drew global attention after its low-cost models matched leading US systems on several benchmarks. A Chinese embassy spokesperson in Washington rejected the security claims as politically motivated, a position Beijing has repeated in prior technology disputes.

The expanded guidance deepened the US-China technology divide and added pressure on companies that embedded DeepSeek's open-weight models into commercial products. The Information Technology Industry Council said it was reviewing the advisory and seeking clarity on compliance timelines.