EU's strict AI law raises the bar for global regulation

https://arab.news/yqkcy Aug. 2, the date when the EU's Artificial Intelligence Act came into full effect this month, may be the most important moment yet in AI legal history. While Europe has many challenges, one area where the 27-member bloc remains a key global force, including influencing organizations across the Middle East, is the global regulatory agenda. To be sure, the new European legislation is by no means the first AI regulation, but it is the most comprehensive. The law represents the strictest set of measures anywhere for the technology. Moreover, organizations failing to comply with rules face fines of potentially as much as €35 million (SR153 million), or 7 percent of global revenue. The regulation's reach is extraterritorial. Specifically, the scope extends to any organization, including those headquartered in the Middle East, whose AI systems are utilized in the EU or whose outputs affect citizens or organizations in the 27 member states. Moreover, many organizations which follow the EU's rules may also likely extend some of those obligations to markets outside the bloc. After all, it is often inefficient to develop separate models for multiple different markets. This is, therefore, the latest example of the so-called 'Brussels effect' of Europe developing legislation that becomes a de facto global standard. By developing far-reaching regulations that have such a significant bearing on the global landscape, the EU helps to ensure that European values shape policy in a wide spectrum of areas from sustainability and environment to antitrust, data privacy plus consumer health and safety. The new European legislation is by no means the first AI regulation, but it is the most comprehensive. Andrew Hammond The worldwide influence that the EU AI regulation already is having is underlined in new insights collated by the Thomson Reuters Foundation's AI Company Data Initiative. This leverages more than 100,000 data points from almost 3,000 organizations. The AI Company Data Initiative database indicates that about 47 percent of organizations citing the European regulation in governance disclosures are not headquartered in the EU, although many have a significant EU market presence. From an industry vantagepoint, the tech sector leads the way accounting for nearly 40 percent of non-EU firms that cite the EU legislation. Meanwhile, communication services and financial services total a combined 29 percent. In the US specifically, the figure for tech firms rises to 53 percent of companies. This reflects the fact that leading players such as Google, Microsoft, xAI, and OpenAI have, so far, voluntarily chosen to adopt key parts of the EU's AI framework, to try to keep access to the 27-member state market. When the EU first approved the legislation in 2024, then-Internal Market Commissioner Thierry Breton claimed that 'the AI Act is a rulebook, but it's also a launchpad for EU startups and researchers to lead the global AI race.' Since then, however, Europe has continued to struggle with the commercialization of AI and the wider tech agenda, certainly compared with other powers such as the US and China, which has led to much criticism. This core fact was showcased perhaps most powerfully in the competitiveness report of former European Central Bank chief Mario Draghi. This highlighted how much the EU's ambition to strengthen its role as a tech leader is being undermined by trends of increased regulation. The EU Act is the latest example of this. Its obligations were phased in earlier this month, seeing all remaining provisions kick in, including rules for systems seen as high-risk. The foundation of the regulation is a risk-based, tiered system where the highest level of regulation applies to those machines that pose the greatest risk to health, safety, and human rights. The highest risk category is now defined by the number of computer transactions needed to train a machine, known as floating point operations per second. The lower tiers of regulation also place major new obligations on AI services. This includes basic rules about disclosure of data used to teach the machine to do anything from writing a newspaper article to diagnosing illness. There will also be strict restrictions on the use of facial recognition technology, except for narrowly defined law enforcement exceptions. Plus bans too on the use of AI for 'social scoring' — using metrics to establish how virtuous someone is — and AI systems that 'manipulate human behavior to circumvent their free will.' The use of AI to exploit those vulnerable because of their age, disability or economic situation will also be banned. Moreover, consumers will have the right to launch complaints and fines could be imposed for violations. Taken together, as much as the AI Act is, therefore, reinforcing the 'Brussels effect,' early evidence indicates it may be doing little to help the bloc fulfill wider goals of boosting tech-led innovation and productivity. With a new digital era dawning, the EU urgently needs to also double down on raising its game, beyond legislation, to improve the bloc's global competitiveness. • Andrew Hammond is an associate at LSE IDEAS at the London School of Economics.