The United Nations on Thursday adopted a U.S.-led resolution on artificial intelligence, marking what Washington says is a major step toward establishing a global baseline to regulate the rapidly developing technology.
The United Nations on Thursday adopted a U.S.-led resolution on artificial intelligence, marking what Washington says is a major step toward establishing a global baseline to regulate the rapidly developing technology.
The resolution, which followed more than three months of negotiations among dozens of countries, calls on U.N. member states to ensure “safe, secure, and trustworthy AI systems” that are developed responsibly and respect human rights and international law.
While the resolution is non-binding and does not include an enforcement mechanism, U.S. officials in a briefing on Wednesday highlighted the significance of its unanimous adoption as an important step in establishing global AI guardrails.
“This first-ever standalone resolution on AI at the United Nations is a consensus resolution—that means that all 193 member states will agree to it, and trust me, that is no easy feat,” a senior Biden administration official said, adding that as of Wednesday afternoon, 97 countries had also co-sponsored the resolution and that number was growing “literally by the hour.”
Debates on how best to regulate AI have dominated bilateral and multilateral forums for more than a year, ranging from the G-7 summit in Japan to the AI Safety Summit hosted by the United Kingdom last November. Several of the world’s most powerful governments have also established their own paths to regulate AI—the European Union earlier this month passed the EU AI Act after nearly two years of deliberations, while authorities in China have cast an ever-expanding, ever-evolving regulatory net to rein in AI technologies.
The Biden administration took its biggest swing last October with an executive order that echoes many of the goals included in the U.N. resolution. “What we’ve done, essentially, is to make sure that the resolution reflects what the administration is already doing with respect to its domestic AI governance,” another senior administration official told reporters.
The United Nations also has multiple other initiatives, including a new AI advisory body and its global standard-setting organization, the International Telecommunication Union. Those efforts will continue, but this week’s resolution may give the conversation more heft. “We view this as complementing other initiatives happening throughout the U.N. system, but it is different,” the second official said. “We think it’s important when all 193 member states agree to a set of global norms.”
That broad agreement is significant, given the diplomatic battles that have played out in the United Nations between Western democracies and allies on the one hand and autocracies on the other. China and Russia, in particular, have increasingly sought to shape the institution toward their worldview and priorities, stalling deliberations over a proposed treaty on crimes against humanity and attempting to impose a contentious treaty on cybercrime. On AI, however, the discussions appear to have been more productive.
“There were lots of heated conversations; that’s not unusual for the United Nations,” the first administration official said. “The fact that 193 countries that often can’t agree on anything at the U.N. were able to agree on this shows that this issue of AI is so transformative—not only from the technology standpoint but in terms of the potential opportunities that people see—that I think it transcended the usual geopolitical divisions that we have here in the United Nations.”
The inclusion of language ensuring AI systems comply with human rights is a particular bright spot of the resolution, according to Daniel Leufer, a senior policy analyst at the digital rights group Access Now. “I wouldn’t take that for granted as a statement,” he said. “Getting the message across that there are uses of AI that are just incompatible with human rights and cannot be permitted was a battle, and it is good to see that enshrined in something at this level with the level of consensus.”
But achieving that consensus also dilutes the impact that the resolution can have, Leufer added, particularly with a lack of enforcement mechanisms built into the U.N. process. “There’s always a risk that what that means effectively is bringing everyone down to the lowest agreeable bar,” he said. “If we limit ourselves to what we can get every state to agree on, we’re not going to get too far.”
One notable absence from the resolution is the potential military use of AI, and that was largely by design. “In looking across the broad sweep of AI considerations in the world, we made a purposeful choice in pursuing a consensus-based U.N. resolution to not include the military uses discussion in this resolution,” one of the officials said, adding that several diplomatic and multilateral conversations about military applications of AI are already ongoing across the U.N. and other forums. “We believed there was an opportunity to talk about safe, secure, and trustworthy AI in a civilian, non-military context, which was very important and deserved and merited its own attention and focus.”
The post The U.N. Gets the World to Agree on AI Safety appeared first on Foreign Policy.