Vice President Kamala Harris said the Biden administration is imposing new binding requirements on federal agencies’ use of artificial intelligence, intending to improve safety, transparency and oversight of new tech tools.
The government-wide policy from the White House Office of Management and Budget announced on Thursday reveals the federal government’s priorities for AI governance as it works to implement President Biden’s AI executive order issued last year.
Ms. Harris told reporters that the Biden administration’s new rules mean government agencies will need to verify that their usage of AI does not endanger Americans’ rights and safety.
“If the Veterans Administration wants to use AI in VA hospitals to help doctors diagnose patients, they would first have to demonstrate that AI does not produce racially biased diagnoses, so that I offer as an example,” Ms. Harris said.
Questions persist about how the Biden administration intends to verify AI tools will not cause harm. Asked about the verification process on Wednesday, a senior Biden administration official told reporters the AI systems will undergo very rigorous testing without providing details of how that would be accomplished.
A White House fact sheet said the new guidance means that airport travelers will be able to opt out of the Transportation Security Administration’s facial recognition technology without any delay or losing their place in line.
Passengers seeing the TSA’s new technology in action at Ronald Reagan Washington National Airport have experienced a different reality.
Sen. Jeff Merkley, Oregon Democrat, attempted to opt out of the facial recognition technology last year at that airport in Virginia and was pressured by a TSA agent telling the senator to stand aside while other passengers were allowed to continue onward.
“I started to step forward and give my ID and he said, ‘Stand aside, you wait here,’ ” Mr. Merkley said in a video published to his Senate website showing the TSA agent’s actions. “And so I did, and in the end, was able to pass through just fine.”
More information about how people’s data is used by the federal government could be forthcoming.
Ms. Harris said a second binding requirement involving transparency means federal agencies will need to publish a list of AI systems that the agencies are using and make it available online each year. Alongside identifying the systems, Ms. Harris said the agencies will need to publish risk assessments that detail how the government is managing danger.
The full picture of the federal government’s use of AI is unlikely to become publicly available, however, as national security and law enforcement agencies often refrain from publishing details of their AI tools amid concerns about exposing their capabilities to foreign adversaries.
Ms. Harris said a third requirement for federal agencies involving internal oversight directs all federal agencies to designate a chief AI officer. Existing AI chiefs are already meeting regularly with the OMB and the White House Office of Science and Technology Policy to prepare for the Biden administration’s AI makeover, according to a White House fact sheet.
“These three new requirements have been shaped in consultation with leaders from across the public and private sector from computer scientists to civil rights leaders to legal scholars and business leaders,” Ms. Harris said. “President Biden and I intend that these domestic policies will serve as a model for global action.”