I advised the EU on its landmark AI Act. Here’s how it got preempted by the U.S. and U.K., why it almost got derailed by ChatGPT, and how it will shape regulation in 2024
For all the talk about a generative artificial intelligence (AI) arms race among the large language model (LLM) developers and Big Tech, it’s policymakers in the U.S. and Europe who are in full sprint mode as they rapidly regulate AI. The introduction of the White House executive order, coinciding with the U.K.’s AI Safety Summit, took much of the AI world by surprise and leveled the balance of tech regulatory power, an area long dominated by the EU.
Now, the EU is taking back its leading regulatory spot as its landmark AI Act on the development and deployment of generative AI takes shape. And it’s bolder and more comprehensive than any of its global counterparts.
As an advisor to the EU as it develops its act, I’ve had a front-row seat to the contentious debates and decisions that have happened behind closed doors over the last few years. Here’s what I know.
Generative AI nearly undid the EU AI Act
The EU AI Act had been in motion for more than four years before the mainstream introduction of LLMs in late 2022. Most AI Act discussions focused on narrow and predictive AI applications, aiming to address the most pressing challenges in these traditional AI applications. How can we regulate AI for diagnostics applications? How do we ensure that evaluations of creditworthiness can be trusted? Amid endless regulatory discussions on what AI actually means, we saw exponential technology advancement unfold right in front of us.
With the introduction of generative AI, the AI Act found itself over a cliff hanging on by a thread. EU policymakers understood the existential risk of simply being outpaced by technology. They faced a big question: Should they hang on to their narrow regulation, or jump into the fray to embrace the new world of generative AI?
Ultimately, they did a bit of both. From my conversations in Brussels over the last months, the divide among policymakers became astoundingly visible. The EU AI Act is a comprehensive document that, at least in terms of its material, is still very much focused on its foundational intent.
What most of us consider the future of AI–and where we will see the most economic activity over the coming years–is addressed through a comparatively small addendum to the original framework. As such, a lot of the fine print for a regulated Generative AI future is yet to be defined.
A lot of technical fine print is yet to come
The act is far from final, with a lot of technical details still to be decided. These fine details are where the regulation will become consequential for LLM developers and enterprises alike. While a foundation framework–consisting of critical yet high-level pieces around transparency requirements and punitive measures–has been set, the act now moves toward robust technical scrutiny from regulators, sectoral agencies, and industry bodies.
This next phase will introduce industry-specific controls and guidelines, shaping how generative AI applications will come to life across verticals. A large task will be to integrate the AI Act proposals with existing regulatory frameworks. It’s in this work that regulators and industry stakeholders will have to show that the Act can move from paper into practice and enforcement.
Whether we like it or not, the EU AI Act will be enforced within the next 12-24 months and companies of all sizes will have to move quickly to get their AI compliance in order.
A recent report found that 59% of developers have security concerns around building and deploying generative AI. Businesses need to equip themselves with both longstanding and emerging capabilities to mitigate these risks.
Above all, the introduction of the EU AI Act is a critical moment for businesses to take an honest look at their AI aspirations and technical and compliance posture.
David Haber is the CEO of Lakera.
More must-read commentary published by Fortune:
- Economic pessimists’ bet on a 2023 recession failed. Why are they doubling down in 2024?
- COVID-19 v. Flu: A ‘much more serious threat,’ new study into long-term risks concludes
- Access to modern stoves could be a game-changer for Africa’s economic development–and help cut the equivalent of the carbon dioxide emitted by the world’s planes and ships
- The U.S.-led digital trade world order is under attack–by the U.S.
The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.
