The office is tasked with overseeing the implementation of the AI Act across the EU while also helping lawmakers find harmony between AI innovation and regulation.
The EU officially opened its new AI Office on Sunday, part of its ongoing plan to establish itself as a global leader in AI policy.
Part of the European Commission (EC), the EU’s executive branch, the AI Office’s creation is the latest step in implementing the AI Act, a piece of legislation set to go into effect this summer that establishes new safety guardrails around certain AI programs in the EU.
In addition to helping to guide the law’s adoption across the EU’s 27 member states, the AI Office will be tasked with overseeing safety evaluations for what it describes as “general-purpose” AI models – though the EC has not yet clarified the parameters of what constitutes “general-purpose.”
Originally drafted in late 2021 and passed in March of this year, the AI Act applies to AI-powered products and services that handle information deemed particularly sensitive or potentially dangerous, such as biometric data. The legislation also requires AI chatbots, such as OpenAI’s ChatGPT, to include clear disclaimers notifying users that they’re interacting with a machine.
According to EC spokesperson Thomas Regnier, roughly 85% of generally available AI products and services will not be regulated under the Act “because they pose low to absolutely no risk in the EU.” Any AI company offering products or services in the EU that falls into that remaining 15% – even if they’re based elsewhere – will be required to comply.
The big challenge, in Regnier’s view, is a perennial one for modern democracies: achieving equilibrium between liberty and security.
“You don’t want to overregulate these technologies,” he said. “We want to have European AI and strong European AI … We’re trying to strike the right balance here between regulation and risks, [while] also investing in this technology, which is the future of all of us, our society and our businesses.”
Across the pond, lawmakers in the US are grappling with the same challenge. The US, along with China, has become a global hub for AI research, innovation and talent. And while some early steps have been taken to regulate AI – including an executive order from the Biden administration aimed at increasing federal oversight of AI model safety testing – nothing as comprehensive as the AI Act has begun to take shape on this side of the Atlantic.
One of the primary reasons for this stagnation, according to Andrew Grotto, a William J Perry international security fellow at Stanford University, is that the US lacks the kind of broad legislative foundation that has enabled the passage of tech-related legislation in Europe in recent years, such as the AI Act, the Digital Markets Act and its sister the Digital Services Act. That foundation is the bloc’s sweeping General Data Protection Regulation, which provides a broad framework for ensuring citizens’ digital privacy. The US, meanwhile, has struggled for decades to meaningfully adapt its digital rights regulations in response to the rise of the internet and social media.
“The chances of Congress passing a comprehensive regulatory framework for AI anytime soon are slim to none,” Grotto says. “I would argue that you probably can’t regulate AI without having some of these underlying policy issues – like privacy rights – worked out.”
Suggested newsletters for you
The EU’s new AI Office will eventually employ more than 140 staffers, including AI specialists, attorneys, policy experts and economists. Of those, around 80 roles have yet to be filled, according to Reigner. The AI Office will have to compete with private companies to recruit top AI talent, an increasingly valuable commodity as deep-pocketed tech firms race to commercialize AI products.
Grotto says this early hiring stage at the AI Office can serve as a bellwether for its orientation as it sets out on a mission to balance regulation with innovation. “In six months’ time,” he says, “if the AI Office has hired more lawyers than it has technologists, that would be an ominous sign.”
For more on the latest happenings in AI, web3 and other cutting-edge technologies, sign up for The Emerging Tech Briefing newsletter.