The Biden administration issued its first set of government-wide artificial intelligence policies. In a press briefing ahead of the announcement, Vice President Kamala Harris said agencies must independently vet their artificial intelligence and share findings with the public. The rules, issued by White House Office of Management and Budget, extend only to tech that could impact the public’s rights or safety. For agencies regulating and administering health care that means tools that help make diagnoses, detect mental health issues, flag patients for interventions and perform other medical administrative functions like health insurance risk assessments. During the briefing, Harris used Veterans Affairs as an example. “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,” she said. Harris outlined three main pillars: — AI tools must not endanger the rights and safety of the American people. — Agencies must publish an annual report revealing where and how they use AI systems, including potential risks and safeguards. — Federal agencies must designate a chief AI officer who can oversee all AI used throughout the agency. Why it matters: Veterans Affairs is already testing AI in its medical centers. In 2019, the agency created the National Artificial Intelligence Institute to test and deploy AI to improve care. Additionally, some VA doctors use AI like Medtronic’s GI Genius, which helps detect colorectal cancer in imaging, and a suicide prevention algorithm. Lawmakers are concerned the VA doesn’t disclose its use of AI to patients. In February, during a hearing on AI before the House Veterans’ Affairs Technology Modernization Subcommittee, Rep. Matt Rosendale (R-Mont.) accused the VA of not disclosing its AI use to patients. “You’re not giving them a choice, and that is dangerous,” he said. Now, the agency will have to give them a choice. The rules require agencies to disclose the AI they use and its risks and performance, and to give people the ability to opt out of using it. What’s next: Government agencies have 60 days to select a chief AI officer and convene an agency governance board to set out a strategy. They have until Dec. 1 to implement risk-management strategies for new and existing AI, though they can ask for an extension.
|