Building guardrails or stifling innovation? Biden’s AI plan tries to split the gap
As the internet and social media grew into massive components of the tech industry and US economy, the federal government has generally kept its hands off.
But the Biden administration believes that artificial intelligence is a different beast. So the president on Monday issued a sweeping executive order that’s designed to make the federal government a major player in the development and deployment of AI systems that can mimic or perhaps someday surpass the creative abilities of human beings.
Federal agencies will assert the right to oversee the development of future AI systems, to minimize their risk to national security and public safety. Meanwhile, government agencies will be tasked to set new standards for AI, aimed at protecting privacy, fending off fraud, and ensuring that AI systems don’t reinforce human prejudices.
The Biden administration has launched an aggressive crackdown on Big Tech, including antitrust lawsuits against Google and Amazon. But the executive order represents the first major attempt in the United States to regulate AI systems on a federal level. And for some industry watchers, it’s long overdue.
“There’s a lot to like in the executive order,” said Gary Marcus, a tech entrepreneur, researcher, and chief executive of the Center for the Advancement of Trustworthy AI. “It’s fantastic that the US government is taking the many risks of AI seriously.”
But Adam Thierer, senior research fellow at the R Street Institute, a conservative think tank, believes the Biden administration’s approach could threaten US competitiveness in AI. Under the executive order, “you have to ask for a permission slip in order to innovate,” said
Thierer. “It undermines our innovation culture and makes innovators and inventors fearful of what will happen if they take risk.” The most aggressive aspect of Biden’s order will seek to force developers of AI “that poses a serious risk to national security, national economic security, or national public health and safety” to report their activities to federal regulators. Such developers must also share the results of any safety tests they conduct on the AI systems. In effect, regulators will have to be told about all potential problems in the AI model during every phase of its development.
The administration said the Defense Production Act gives it the authority to enforce the order. This law, passed in 1950 during the Korean War, empowers the US government to make high-priority demands of private businesses in cases of national emergency. The law was invoked under the Trump and Biden administrations to speed up production of vaccines and other medical supplies during the COVID-19 pandemic.
But Thierer said the administration is exceeding its authority and abusing the 1950 law by “reading into it a blank check for government to regulate AI aggressively. It’s just not there.”
Marcus, while supportive of the order, also raised concerns: “Does this [order] apply only to defense procurements, or to any consumer product that could pose safety threats?” In addition, he said, “who decides what poses a serious risk, and how?“Marcus said companies may well challenge the order in court.
Other portions of the executive order seek to provide guidance to corporations and government agencies on how best to use AI systems. The National Institute of Standards and Technology is ordered to set new safety testing standards for AI systems to be used in critical sectors like hospitals, transportation networks, and public utilities. AI systems used in biotechnology research will be designed to prevent their being used to engineer biological weapons. The Department of Commerce will set standards for “watermarking” AI-generated images, text, and other content to prevent its use in fraudulent documents or “deep fake” images.
The order also calls for new efforts to develop AI systems that can better protect consumer privacy, and prevent the use of AI systems that contain racial, gender, or other forms of bias.
Thierer said many of the issues addressed by Biden’s order should be the province of Congress. But lawmakers have yet to introduce an AI bill.
“AI regulation by executive order is what we’re stuck with, because of congressional dysfunction,” said Thierer.
Max Tegmark, a physics professor at the Massachusetts Institute of Technology and a prominent critic of the AI industry, praised the executive order but urged even stronger action, in an email Monday.
“The United States is already far behind Europe in developing and implementing policies that can make technological innovation sustainable by reducing the threats and harms presented by out-of-control, unchecked AI development,” said Tegmark, who earlier this year was among the scientists and tech leaders calling for a six-month moratorium on AI development out of concern the technology threatened public safety. (Tegmark is president of the nonprofit Future of Life Institute, which published the petition.)
“Policymakers, including those in Congress, need to look out for their citizens by enacting laws with teeth that tackle threats and safeguard progress,” Tegmark wrote this week.