AI industry frets about U.S. executive order’s impact on data sovereignty

U.S. President Joe Biden is signing an executive order on artificial intelligence (AI). [Photo by Washington DC EPA / Yonhap]

With U.S. President Joe Biden recently signing an executive order on artificial intelligence (AI) that requires AI developers to share their information with the U.S. government, South Korean AI companies have expressed concerns over the potential impacts the decision will have on their data sovereignty.

Biden issued a landmark executive order to mitigate AI risks on Monday, which laid out new standards for AI safety and security to protect national security, consumers, workers, and minority groups, according to a statement published by the White House.

Citing the Defense Production Act, the order requires companies developing any AI model that poses a severe threat to national security, national economic security, or national public health and safety to share their safety test results and other critical information with the U.S. government. Under the executive order, the National Institute of Standards and Safety (NIST) will be responsible for developing standards for “red team” AI models before their public release.

The U.S. government will also issue guidance with labeling and watermarking for AI-generated content to help differentiate between authentic interactions and those generated by software. Before the order’s issuance, 15 global AI developers, including OpenAI and Microsoft, made voluntary commitments to safety standards by promising to add watermarks to AI-generated content.

“To realize the promise of AI and avoid the risk, we need to govern this technology,” Biden said. “In the wrong hands, AI can make it easier for hackers to exploit vulnerabilities in the software on which our society runs.”

But the Korean software industry voiced concerns over the executive order, under which the U.S. government is authorized to collect a significant portion of data and information from AI developers. The Commerce Department will require companies developing or demonstrating an intent to develop potential dual-use foundation models to report certain information and activities to the federal government within 90 days. Companies must also implement physical and cyber security safeguards to ensure data integrity.

The U.S. government included any model that was trained using a quantity of computing power greater than 1026 integer or floating-point operations in the regulatory scope, which has triggered concerns about the stringent censorship against AI developers.

“The U.S. government’s recent action appears to establish a global barrier that provides control over foreign AI tech firms, potentially obliging them to share data with the U.S. government if they want to launch services in the country,” an unnamed representative from an AI software enterprise said.