Washington removes Chinese DeepSearch from its civilian agencies and bans software built on its models

The U.S. Department of Homeland Security issued an order requiring civilian federal agencies to remove all artificial intelligence technologies from the Chinese company "DeepSearch," an action that began to be implemented in late July before it was publicly disclosed on Monday. The decision covered major federal agencies, including the Treasury, State and Justice departments, with an explicit exemption for entities under the Department of Defense. Although the official notice did not detail the specific security concerns, the move aligns with the approach of the administration of President Donald Trump, who emphasized that China is the primary competitor in the race for technological advancement.

The most sensitive impact of the decision is not limited to banning the model itself; it also extends to prohibiting any software or technology created with the assistance of DeepSearch models.The move followed an internal memo written by the White House technology official Michael Kratzyus, in which he warned that U.S. companies could be subject to harmful exploitation by competing Chinese entities. It coincided with accusations from OpenAI and Anthropic, which are preparing to go public in the coming months, that their models have been subject to cloning by Chinese firms, including DeepSearch.

The official ban comes as DeepSearch is experiencing rapid growth, with its free app topping download charts in the Apple Store and American platforms such as Snowflake, Verse and OpenRouter incorporating its models into developer services. The main driver of this momentum is the large cost gap: the Chinese company said that training its first model cost about $6 million, compared with roughly $100 million to train GPT-4, which at the time caused tech-stock declines. The company reinforced its presence by launching its latest version last week, which received positive reviews for its superiority in complex code-writing tasks.

For development teams and startups in the Gulf, Egypt and the Levant, the decision reshapes the viability of relying on low-cost open-source models.DeepSearch has represented an attractive economic option for reducing computing and inference costs compared with closed models such as Cloud and Gemini. However, the ban’s scope covering derived technologies built with these models means that any regional software product that relies on DeepSearch weights could become ineligible for contracts or integration with international platforms that operate within the civilian supply chains of the U.S. government.

Choosing an AI architecture in regional projects is no longer just a pricing issue. If you are building software solutions aimed at expanding into markets subject to U.S. compliance, leveraging low-cost models will require early scrutiny of the code-building chain to avoid later mandatory modification burdens.