President Joe Biden on Monday signed a wide-ranging executive order on artificial intelligence, covering topics as varied as national security, consumer privacy, civil rights and commercial competition. The administration heralded the order as taking "vital steps forward in the U.S.'s approach on safe, secure, and trustworthy AI."
The order directs departments and agencies across the U.S. federal government to develop policies aimed at placing guardrails alongside an industry that is developing newer and more powerful systems at a pace rate that has many concerned it will outstrip effective regulation.
"To realize the promise of AI and avoid the risk, we need to govern this technology," Biden said during a signing ceremony at the White House. The order, he added, is "the most significant action any government anywhere in the world has ever taken on AI safety, security and trust."
One of the marquee requirements of the new order is that it will require companies developing advanced artificial intelligence systems to conduct rigorous testing of their products to ensure that bad actors cannot use them for nefarious purposes. The process, known as red teaming, will assess, among other things, "AI systems threats to critical infrastructure, as well as chemical, biological, radiological, nuclear and cybersecurity risks."
The National Institute of Standards and Technology will set the standards for such testing, and AI companies will be required to report their results to the federal government prior to releasing new products to the public. The Departments of Homeland Security and Energy will be closely involved in the assessment of threats to vital infrastructure.
To counter the threat that AI will enable the creation and dissemination of false and misleading information, including computer-generated images and "deep fake" videos, the Commerce Department will develop guidance for the creation of standards that will allow computer-generated content to be easily identified, a process commonly called "watermarking."
The order directs the White House chief of staff and the National Security Council to develop a set of guidelines for the responsible and ethical use of AI systems by the U.S. national defense and intelligence agencies.
The order proposes a number of steps meant to increase Americans' privacy protections when AI systems access information about them. That includes supporting the development of privacy-protecting technologies such as cryptography and creating rules for how government agencies handle data containing citizens' personally identifiable information.
However, the order also notes that the United States is currently in need of legislation that codifies the kinds of data privacy protections that Americans are entitled to. Currently, the U.S. lags far behind Europe in the development of such rules, and the order calls on Congress to "pass bipartisan data privacy legislation to protect all Americans, especially kids."
The order recognizes that the algorithms that enable AI to process information and answer users' questions can themselves be biased in ways that disadvantage members of minority groups and others often subject to discrimination. It therefore calls for the creation of rules and best practices addressing the use of AI in a variety of areas, including the criminal justice system, health care system and housing market.
The order covers several other areas, promising action on protecting Americans whose jobs may be affected by the adoption of AI technology; maintaining the United States' market leadership in the creation of AI systems; and assuring that the federal government develops and follows rules for its own adoption of AI systems.
Experts say that despite the broad sweep of the executive order, much remains unclear about how the Biden administration will approach the regulations of AI in practice.
Benjamin Boudreaux, a policy researcher at the RAND Corporation, told VOA that while it is clear the administration is "trying to really wrap their arms around the full suite of AI challenges and risks," much work remains to be done.
"The devil is in the details here about what funding and resources go to executive branch agencies to actually enact many of these recommendations, and just what models a lot of the norms and recommendations suggested here will apply to," Boudreaux said.
Looking internationally, the order says the administration will work to take the lead in developing "an effort to establish robust international frameworks for harnessing AI's benefits and managing its risks and ensuring safety."