House staffers are being urged not to use the Chinese artificial intelligence program DeepSeek, which has soared in popularity since its recent launch.
In a notice to staffers obtained by POLITICO, the House Chief Administrator Officer said it is reviewing the technology, and until then it isn’t authorized for official House use.
That means staffers can’t install it on “House-issued devices, including phones, computers and tablets,” and extra precautions have been taken to restrict its functionality on those devices.
“Threat actors are already exploiting DeepSeek to deliver malicious software and infect devices,” reads the CAO notice.
DeepSeek’s surprise emergence as an AI success story for China has led to bipartisan calls for a crackdown on sending high-tech microchips to the country. Key lawmakers have declared it a wake-up call for the U.S. to ramp up its race against China in the continued development of AI technology.
Speaker Mike Johnson appeared to warn against use of DeepSeek during a press briefing Monday at the House GOP retreat in Florida, saying China “abuse[s] the system, they steal our intellectual property.”
He added: “They’re now trying to get a leg up on us in AI.”
Hours later, however, President Donald Trump argued the technology may be “a positive development.”
House staffers — granted anonymity to share private concerns that may or may not be shared by the lawmakers they work for — noted the rare step of restricting the functionality of any program on House devices, though the House has stepped up such moves in recent years. It similarly banned staffers from using TikTok on official congressional devices last year over security concerns related to its Chinese parent company, ByteDance.
In August 2024 the House Chief Administrative Officer and House Administration Committee also launched a new policy for the safe deployment of artificial intelligence tools by member and institutional offices.
Under the approval process, the CAO reviews AI tools for “defined use cases” and the House Administration panel approves them. That process is intended to help “offices to reduce the security and privacy risks associated with AI,” according to the CAO overview’s overview.
Katherine Tully-McManus contributed to this report.
