The European Union unveiled strict regulations on Wednesday to govern the use of artificial intelligence, a first-of-its-kind policy that outlines how companies and governments can use a technology seen as one of the most significant, but ethically fraught, scientific breakthroughs in recent memory.
The draft rules would set limits around the use of artificial intelligence in a range of activities, from self-driving cars to hiring decisions, bank lending, school enrollment selections and the scoring of exams. It would also cover the use of artificial intelligence by law enforcement and court systems — areas considered “high risk” because they could threaten people’s safety or fundamental rights.
Some uses would be banned altogether, including live facial recognition in public spaces, though there would be several exemptions for national security and other purposes.
The 108-page policy is an attempt to regulate an emerging technology before it becomes mainstream. The rules have far-reaching implications for major technology companies including Amazon, Google, Facebook and Microsoft that have poured resources into developing artificial intelligence, but also scores of other companies that use the software to develop medicine, underwrite insurance policies, and judge credit worthiness. Governments have used versions of the technology in criminal justice and allocating public services like income support.
Companies that violate the new regulations, which could take several years to move through the European Union policymaking process, could face fines of up to 6 percent of global sales.
“On artificial intelligence, trust is a must, not a nice to have,” Margrethe Vestager, the European Commission executive vice president who oversees digital policy for the 27-nation bloc, said in a statement. “With these landmark rules, the E.U. is spearheading the development of new global norms to make sure A.I. can be trusted.”
The European Union regulations would require companies providing artificial intelligence in high-risk areas to provide regulators with proof of its safety, including risk assessments and documentation explaining how the technology is making decisions. The companies must also guarantee human oversight in how the systems are created and used.
Some applications, like chatbots that provide humanlike conversation in customer service situations, and software that creates hard-to-detect manipulated images like “deepfakes” would have to make clear to users that what they are seeing is computer generated.
For the past decade, the European Union has been the world’s most aggressive watchdog of the technology industry, with its policies often used as blueprints by other nations. The bloc has already enacted the world’s most far-reaching data-privacy regulations, and is debating additional antitrust and content-moderation laws.
But Europe is no longer alone in pushing for tougher oversight. The largest technology companies are now facing a broader reckoning from governments around the world, each with their own political and policy motivations, to crimp the industry’s power.
In the United States, President Biden has filled his administration with industry critics. Britain is creating a tech regulator to police the industry. India is tightening oversight of social media. China has taken aim at domestic tech giants like Alibaba and Tencent.
The outcomes in the coming years could reshape how the global internet works and how new technologies are used, with people having access to different content, digital services or online freedoms based on where they are located.
Artificial intelligence — where machines are trained to perform jobs and make decisions on their own by studying huge volumes of data — is seen by technologists, business leaders and government officials as one of the world’s most transformative technologies, promising major gains in productivity.
But as the systems become more sophisticated it can be harder to understand why the software is making a decision, a problem that could get worse as computers become more powerful. Researchers have raised ethical questions about its use, suggesting that it could perpetuate existing biases in society, invade privacy, or result in more jobs being automated.
Release of the draft law by the European Commission, the bloc’s executive body, drew a mixed reaction. Many industry groups expressed relief the regulations were not more stringent, while civil society groups said they should have gone further.
“There has been a lot of discussion over the last few years about what it would mean to regulate A.I., and the fallback option to date has been to do nothing and wait and see what happens,” said Carly Kind, director of the Ada Lovelace Institute in London, which studies the ethical use of artificial intelligence. “This is the first time any country or regional bloc has tried.”
Ms. Kind said many had concerns the policy was overly broad and left too much discretion to companies and technology developers to regulate themselves.
“If it doesn’t lay down strict red lines and guidelines and very firm boundaries about what is acceptable, it opens up a lot for interpretation,” she said.
The development of fair and ethical artificial intelligence has become one of the most contentious issues in Silicon Valley. In December, the co-leader of a team at Google studying ethical uses of the software said she was fired for criticizing the company’s lack of diversity and the biases built into modern artificial intelligence software. Debates have raged inside Google and other companies about selling the cutting-edge software to governments for military use.
In the United States, the risks of artificial intelligence are also being considered by government authorities.
This week, the Federal Trade Commission warned against the sale of artificial intelligence systems that use racially biased algorithms, or ones that could “deny people employment, housing, credit, insurance, or other benefits.”
Elsewhere, in Massachusetts, and in cities like Oakland, Calif.; Portland, Ore.; and San Francisco, governments have taken steps to restrict police use of facial recognition.