US Congress advances broader AI chip export controls targeting China beyond current restrictions

Washington is facing renewed pressure to tighten chip export controls as lawmakers weigh new measures aimed at preserving the U.S. lead in artificial intelligence. The debate has intensified following concerns that Chinese memory maker CXMT is expanding rapidly while U.S. policy still leaves gaps in restrictions on semiconductor equipment, advanced chips, and remote access to data centers.

The proposed legislation would bar servicing of equipment already sold to China, gradually weakening existing production capacity. The AI Overwatch Act would prohibit sales of advanced chips and require the Commerce Department to ensure foreign shipments do not reduce domestic supply. The Chip Security Act targets chip smuggling directly.

The policy debate intensified amid criticism that the Trump administration has allowed exports of increasingly advanced chips and halted expansion of the Entity List of Chinese companies. Despite being identified by the Pentagon as a Chinese military company, CXMT has not been added to the Commerce Department's Entity List and has received looser treatment than some critics consider appropriate.

CXMT became a major memory chip supplier within a decade, backed by substantial state support—over one-third of its equity is held directly by the Chinese government. PC makers HP, Asus, and Acer already use its chips, while Apple is qualifying them for devices sold in China and seeking approval to add them to its U.S. supply chain.

A significant gap remains in U.S. controls: remote access by Chinese firms to overseas data centers equipped with advanced chips. The Remote Access Security Act passed the House of Representatives but remains stalled in the Senate, creating what critics describe as a critical loophole in Washington's technology competition with China.

Political backlash over AI agent hacks and White House AI oversight has intensified as lawmakers from both parties questioned whether voluntary testing and limited safeguards suffice, while scrutinizing the administration's ties to major AI firms and potential industry influence on policy decisions.