Europe pitches ‘AI Pact’ to curtail the booming tech’s risks

Europe has a new plan to get ahead of dangerous artificial intelligence: An “AI Pact.”

The boom of generative artificial intelligence tools like chatbot ChatGPT has triggered a global scramble of regulators trying to keep the technology under control. In meetings with industry leaders this week, and with global peers in past months, the European Union — the Western world’s de facto digital policeman — has quietly been building support for a global pledge to build and use these new tools responsibly.

“We are in the middle of things,” Margrethe Vestager, the Commission’s executive vice president responsible for digital affairs, told reporters this week. “And while we’re in the middle of things, we see generative AI and the large language models are developing really, really fast.”

At a meeting of G7 countries in Hiroshima, Japan, last weekend, leaders pledged to advance “international discussions on inclusive artificial intelligence (AI) governance and interoperability to achieve our common vision and goal of trustworthy AI.” Next week, European and United States leaders will echo that pledge, calling for assessments of “emergent behaviours and risks” of AIs like ChatGPT, according to a draft statement for their upcoming Trade & Technology Council (TTC) seen by POLITICO.

The European Commission already proposed its Artificial Intelligence Act — establishing binding rules on banned and “high-risk” AI applications — in 2021, but the law is still being reviewed by lawmakers and EU countries and is expected to come into force no earlier than two years from now. EU officials are now trying to shape a voluntary AI pact — one that would see companies like Microsoft, Google and OpenAI agree to nonbinding principles around transparency and accountability — as a stopgap solution for the rapidly developing technology.

Already, the EU’s top officials have found common ground with industry heavyweights.

The rush to regulate AI’s dangers has drawn the industry’s leading executives to Brussels, other European capitals, and even Washington this week. The goal is to demonstrate that the industry is open to regulation — as long as it isn’t too onerous.

Google CEO Sundar Pichai was in Brussels on Wednesday for a series of AI-focused talks with policymakers, including the great and the good from the Commission and European Parliament. The Indian engineer’s pitch was simple: The search giant wants to be a responsible developer of AI, and it welcomes greater guardrails, but Europe’s AI Act would likely hobble the industry from harnessing the technology for good. 

Pichai’s meeting with Vestager and Internal Market Commissioner Thierry Breton triggered statements from the top ranks of the Berlaymont building around the so-called AI Pact. In truth, those words are still premature, after Vestager secured a tentative agreement during a G7 digital meeting last month to work on problems associated with generative AI sometime later this year, in a meeting to be held in Japan, which currently holds the G7 presidency.

While Pichai was roaming around Europe, Sam Altman, the OpenAI chief, traveled to Spain, France, Poland and the United Kingdom, meeting with national leaders Pedro Sánchez, Emmanuel Macron, Mateusz Morawiecki and Rishi Sunak, as well as key ministers, to pitch his own version of AI rulemaking — one that borrowed heavily from testimony he gave to the U.S. Senate earlier in the month.

Altman’s pitch was at once different from and similar to Pichai’s. 

Ahead of his tour, he called for the establishment of a global organization — modeled after the International Atomic Energy Agency — to stave off existential risks posed by smarter-than-human AIs. Altman said current AI models don’t need to be regulated too strictly — a direct attack on Europe’s own AI proposals.

Following reports that Altman was considering pulling ChatGPT out of the EU, he appeared to backtrack, saying Friday that “we plan to comply [with EU rules]. I really want to offer services in Europe. We just want to make sure we’re technically able to. And I think the conversations have been super productive this week.”

In Washington, Microsoft’s President Brad Smith made a similar pitch to lawmakers in an effort to get ahead of any potential rulemaking that may come from the U.S. Capitol. The tech giant, which has invested heavily in OpenAI, is hoping to use generative AI to take on Google’s dominance in search. 

Speaking in front of a gaggle of D.C. lawmakers, he urged greater regulation on this fast-moving technology, including cybersecurity checks for critical infrastructure and a licensing regime for AI models. 

“I’m not even sure that we’re in the car,” Smith told reporters after rejecting the notion that Microsoft was in the driver’s seat on how the technology was developing. “But we do offer points of view and suggested directions for those who are actually driving.”

Laura Kayali, Mark Scott and Aoife White contributed reporting.