By Guardian - Alex Hern - Tue 28 May 2024

One day and six (very long) agreements later, can we call the meeting to hammer out the future of AI regulation a success? What does success look like for the second global AI summit?

As the great and good of the industry (and me) gathered last week at the Korea Institute of Science and Technology, a sprawling hilltop campus in eastern Seoul, that was the question I kept asking myself.

If we’re ranking the event by the quantity of announcements generated, then it’s a roaring success. In less than 24 hours – starting with a virtual “leader’s summit” at 8pm and ending with a joint press conference with the South Korean and British science and technology ministers – I counted no fewer than six agreements, pacts, pledges and statements, all demonstrating the success of the event in getting people around the table to hammer out a deal.

There were the Frontier AI Safety Commitments:

The first 16 companies have signed up to voluntary artificial intelligence safety standards introduced at the Bletchley Park summit, Rishi Sunak has said on the eve of the follow-up event in Seoul.

“These commitments ensure the world’s leading AI companies will provide transparency and accountability on their plans to develop safe AI,” Sunak said. “It sets a precedent for global standards on AI safety that will unlock the benefits of this transformative technology.”

The [deep breath] Seoul Statement of Intent toward International Cooperation on AI Safety Science:

Those institutes will begin sharing information about models, their limitations, capabilities and risks, as well as monitoring specific “AI harms and safety incidents” where they occur and sharing resources to advance global understanding of the science of AI safety.

At the first “full house” meeting of those countries on Wednesday, [Michelle Donelan, the UK technology secretary] warned the creation of the network was only a first step. “We must not rest on our laurels. As the pace of AI development accelerates, we must match that speed with our own efforts if we are to grip the risks and seize the limitless opportunities for our public.”

The Seoul Ministerial Statement:

Twenty-seven nations, including the United Kingdom, Republic of Korea, France, United States, United Arab Emirates, as well as the European Union, have signed up to developing proposals for assessing AI risks over the coming months, in a set of agreements that bring the AI Seoul summit to an end. The Seoul Ministerial Statement sees countries agreeing for the first time to develop shared risk thresholds for frontier AI development and deployment, including agreeing when model capabilities could pose “severe risks” without appropriate mitigations. This could include helping malicious actors to acquire or use chemical or biological weapons, and AI’s ability to evade human oversight, for example by manipulation and deception or autonomous replication and adaptation.

There was also the Seoul Declaration for safe, innovative and inclusive AI, which outlined the common ground on which 11 participating nations and the EU agreed to proceed, and the Seoul Statement of Intent toward International Cooperation on AI Safety Science, which laid out a rough sense of what the goals actually were.

And the Seoul AI Business Pledge, which saw 14 companies – only partially overlapping with the 16 companies that had earlier signed up to the Frontier AI Safety Commitments – “committing to efforts for responsible AI development, advancement, and benefit sharing”.

Call to action

It’s understandable if your eyes glazed over. I’m confused, and I was there.

The issue isn’t helped by the dual hosts of the event. For the UK, the Frontier AI Safety Commitments were the big ticket, announced with a comment from Rishi Sunak and paired with the offer of an interview with the technology secretary for press in Seoul. In the English-language press, the Seoul AI Business Pledge was all but ignored, despite it being a significant plank of the South Korean delegation’s achievements. (My guess, for what it’s worth, is that the focus on “Frontier AI” from the first set of commitments slightly short-changed the South Korean technology sector, including as it did just Samsung and Naver; the Seoul pledge, by contrast, was six domestic firms and eight international ones.)

It might be simpler to explain it all by breaking it down by group. There’s the two competing pledges from businesses, each detailing a slightly different voluntary code they intend to follow; there’s the three overlapping statements from nations, laying out what they actually wanted to get from the AI summit and how they’re going to get there over the next six months; and there’s the one concrete plan of action from the national AI safety institutes, detailing how and when they’re going to work together to understand more about the cutting-edge technology they’re trying to examine.

There’s an obvious objection here, which is that none of these agreements have teeth, or even really sufficient detail to identify whether or not someone is trying to follow them. “Companies determining what is safe and what is dangerous, and voluntarily choosing what to do about that, that’s problematic,” Francine Bennett, the interim director of the Ada Lovelace Institute, told me.

Similarly, it feels weird to hold the agreements up as the success of the summit when they were largely set in stone before delegates even arrived in South Korea. At times, as the agreements and releases continued to hit my inbox with scant relationship to the events happening in person, it felt like a summit that could have been an email.

Just to be invited

The truth is slightly different: the success of the summit is that it happened.

That sounds like the faintest of praise, but it’s true of any event like this. Yes, the key agreements of the summit were all pieced together before it started – but in providing a hard deadline and a stage from which to announce success, the event gave everyone a motivation to sign up.

And while the agreements are certainly toothless, they’re also a starting point for the real work to be done once the summit ends. Companies and governments signing up to a shared description of reality is the first step to being able to have the difficult technical conversations required to fix the problems. “We all want the same thing, here” is a powerful statement when it’s true. And, when you draw the boundaries carefully enough, it still is.