Biden signs AI order calling for beefed-up critical infrastructure protections

Amid growing concerns over cybersecurity threats posed by emerging artificial intelligence technologies, President Joe Biden ordered a broad review of the risks to the nation's critical infrastructure, including energy installations, and ordered the development of new AI standards.

The review was one of a suite of measures included in an executive order issued Oct. 30. A fact sheet released by the White House said the order called on the departments of Energy and Homeland Security to "address AI systems' threats to critical infrastructure, as well as chemical, biological, radiological, nuclear and cybersecurity risks."

The administration also tasked the National Institute of Standards and Technology with developing "rigorous standards for extensive red-team testing to ensure safety before public release" of AI systems. The DHS would be responsible for applying those new standards to critical infrastructure, along with establishing an AI safety and security board.

The infrastructure-related measures were part of a broader set of actions aimed at AI security, including a requirement that developers of "any foundation model that poses a serious risk to national security, national economic security, or national public health and safety" share the results of safety tests with the federal government.

The White House called the executive order "the most significant actions ever taken by any government to advance the field of AI safety."

In public comments made before signing the order, Biden said, "In the wrong hands, AI can make it easier for hackers to exploit vulnerabilities in the software that makes our society run."

Biden said the order would require the DHS and the Defense Department "to develop game-changing cyber protections that will make our computers and our critical infrastructure more secure than it is today."

Improvements lead to greater use

The president's order comes as recent improvements in AI have led many companies to evaluate the widespread deployment of such technologies to improve data analysis and optimize corporate decision-making.

In one such case, Avangrid, an energy conglomerate, said in an interview earlier this year that it is launching an in-house team that will use AI data on the performance of its assets and weather and geospatial data to determine where to focus its grid maintenance investments.

Along with that optimism has come concern that hostile actors could use new AI technologies to harm US interests. China, with its advanced technology industry, has been a key focus for Washington, with policymakers worried that Chinese AI capabilities could soon surpass those of the US.

"Chinese [AI] investments are expected to reach over $26 billion by 2026," dwarfing the $3.3 billion being spent annually by the US, Deputy Energy Secretary David Turk told lawmakers in September. "We simply must be bolder and move faster or risk falling behind."

At the same time, Turk warned that AI "lowers the bar for bad actors to do even worse things and to do those worse things easier."

Beyond the familiar cybersecurity concerns, Turk also warned that AI poses risks to individuals' privacy and civil liberties. More broadly, he said AI could be used to spread misinformation, manipulate markets, and develop biological, nuclear, and chemical threats.

Long history with AI research

The DOE has long been at the forefront of AI research; the department's investments in such technologies date back to the 1960s. In the last decade, thousands of AI applications and four of the top 10 fastest supercomputers have emerged from the department. The DOE's Oak Ridge National Laboratory houses Frontier, the fastest supercomputer in the world, which can perform 1 quintillion -- or 1 billion billion -- calculations per second.

In a statement after the administration's announcement, Tom Quaadman, executive vice president of the US Chamber of Commerce's Technology Engagement Center, said his group "appreciates the priorities outlined in the executive order," such as the development of new AI standards. But he noted that "substantive and process problems still exist" and suggested the order could lead to the imposition of unhelpful federal regulations.

"Short overlapping timelines for agency-required action endangers necessary stakeholder input, thereby creating conditions for ill-informed rulemaking and degrading intra-government cooperation," Quaadman said.

Quaadman also said federal agencies "should not view this as a license to do as they please -- all agencies must continue to act within the limits of their congressional mandates and abide by the major questions doctrine as articulated by the Supreme Court."