Key Points

This legislation, co-sponsored by Representatives Ritchie Torres (D-NY) and Darin LaHood (R-IL) in the House and introduced by Senators Rick Scott (R-FL) and Gary Peters (D-MI) in the Senate, requires that the Federal Acquisition Security Council create and frequently update a list of AI technologies developed by adversarial entities.

DeepSeek serves as an example, having documented connections to the Chinese Communist Party and its intelligence network, and its privacy policy clearly states that user data from the United States is stored in China.

In April, Ranking Member Krishnamoorthi and Chairman Moolenaar unveiled the results of their investigation into DeepSeek, which included a recommendation to Congress to "impose a federal procurement ban on AI models originating from the PRC, including a prohibition on their use on government devices," as articulated in the DSCCCP release.

The No Adversarial AI Act would prohibit US government agencies from acquiring or utilizing AI products created by firms like DeepSeek unless a specific exception is provided for research, testing, or national security needs, requiring notification to Congress.

This legislation also addresses adversarial AI technologies that may arise in the future, which could represent a similar or even more significant threat to federal systems and data, according to the DSCCCP release.

"Artificial intelligence controlled by foreign adversaries represents a direct risk to our national security, data integrity, and government operations," stated Ranking Member Krishnamoorthi.

"We must not permit hostile regimes to embed their software in our most vital systems. This bipartisan measure will establish a clear barrier between foreign adversary AI and the U.S. government, safeguarding our institutions and the American populace. Chinese, Russian, and other adversarial AI systems have no place on government devices and certainly should not be entrusted with government data," as quoted in the DSCCCP release.

"We are currently in a new Cold War, with AI emerging as the strategic technology at its core," remarked Chairman Moolenaar.

"The CCP does not innovate; it steals, scales, and undermines. From intellectual property theft to chip smuggling to the integration of AI into surveillance and military platforms, the Chinese Communist Party is hastening efforts to weaponize this technology. We must draw a definitive line: systems used by the U.S. government cannot be powered by tools designed to serve authoritarian objectives."

"The Communist Chinese regime will resort to any tactic to spy, steal, and weaken the United States, and as AI technology progresses, we must enhance our efforts to protect national security and prevent adversarial regimes from exploiting technology against us," expressed Senator Scott.

"Given the clear evidence that China can access US user data on AI platforms, it is reckless for our federal agencies to utilise these hazardous systems and expose our government to Beijing's influence. Our No Adversarial AI Act will mitigate this direct threat to our national security and ensure that sensitive data remains secure from enemy hands."

"Artificial intelligence holds immense promise for our economy and society, but it also presents real security risks when leveraged by foreign adversaries," said Senator Peters.

"This legislation helps safeguard US government systems from AI developed by foreign adversaries that could compromise our national security or put Americans' data at risk. It's a smart, focused step to ensure our government technology infrastructure keeps pace with the evolving threats we face while still allowing room for scientific research, evaluation, and innovation. I'm proud to support this effort to protect Michiganders' personally identifiable information from bad actors who could exploit their data housed on government systems," the DSCCCP release quoted.

The bill empowers the federal government to identify, exclude, and remove adversarial AI from its systems while enhancing transparency and oversight of federal AI.

- ANI