The European Parliament has approved the AI Act which is a comprehensive framework that attempts to address the risks that AI presents. While countries like the US and China have made some attempt at regulating AI, the EU’s AI Act goes further and is the world’s first set of binding requirements to mitigate AI risk. Enza Iannopolla, principal analyst at Forrester, believes that this would make the EU the standard for trustworthy AI and other countries would attempt to reach the same level (McCallum et al.).

An important consideration for the EU is that while this legislation may be the leading of its kind, the US- followed by China- dominates when it comes to AI. While these regulations may promote trustworthy AI, there is also the risk that it would limit the creation of AI in Europe. The AI Act could create a competitive disadvantage for Europe as companies may be more inclined to develop in regions with less strict regulations and the regulations could limit European companies (Ditsche and Mikhaylenko). On the other hand, the regulations could also aid in increasing Europe’s soft power as the AI Act requires companies with generative AI tools to disclose the material they use to train their AIs (McCallum and Vallance). AI has gotten a bad reputation for using copyrighted material and the Act could improve the perception of AI amongst artists and generate goodwill towards the EU for tackling the problem.

Leader in AI Regulation

The EU wants to promote trustworthy AI and the AI Act allows them to do so while also placing themselves in an influential position which could generate AI activity in Europe. Regulating AI is not a new concept, although its implementation has been slow. For instance, in April, the UK and the US signed a Memorandum of Understanding in order to work together to test advanced AIs (UK Gov). The largest AI firms are cooperating with the idea of regulation, but the fact remains that till now regulators have not implemented any significant restrictions on the companies’ goals or asked for the data they use to train their AI (McMahon and Kleinman).

The AI Act is meant to establish the EU as a leader at the forefront of AI, and it is succeeding as countries like the UK are recognizing the advancements the EU is making. This could be a sore point for the UK; they currently have more funding in AI safety than any other government but they have been overtaken by the EU for risk mitigation regulation (Vallance). Post-Brexit tensions still exist and considering a large motivator of leaving the EU was the ability to take back control, there may be pressure internally for the government to show that they are capable of outperforming the EU. The government wants to use their newfound freedom and create policy, but unless this is exercised in a way that places the UK in a comparable position to the EU when it comes to AI regulation, then perhaps the benefits of Brexit are not to be found in this area (Marshall and Goss).

There was a study that showed that regions who were early in creating ideas in new scientific fields, had an innovation advantage in those fields over time. The study covered fields like AI, and found that the EU was significantly behind the US in terms of their innovation advantage (Filimonovic et al.) Considering the study set out that early leaders in scientific innovation had an advantage in technological innovation, it would perhaps take more than was reasonable for the EU to catch up to the US in the traditional way. By focusing on risk mitigation regulation, the EU is using considerably less resources and taking an unconventional route to lead the market.

World’s Unicorn Capital

The US is leading the field when it comes to AI and the EU is attempting to level it by introducing the AI Act. There are those who believe that the AI act will limit the EU’s advancement and place them at a competitive disadvantage. However, it needs to be kept in mind that the US is also attempting to regulate AIs. For instance, President Biden declared an executive order requiring data sharing from AI developers to the government (McCallum et al.). Senator Wiener in California even authored a bill that would introduce AI regulations but it faced opposition from tech companies and was blocked by Governor Newsom as he believed that it would reduce innovation and developers may move out of the state (Silva).

These criticisms are the same as those that the EU AI Act has received, and they are not completely unfounded. For instance, OpenAI CEO Altman initially made threats on leaving the EU due to the planning of the AI Act as he believed it would be overregulation (McCallum and Vallance). Nevertheless, regulation is important especially considering what AI is capable of. For example, OpenAI has developed a voice cloning tool but made a decision not to release it due to the risks that it presents, particularly in election years. In fact, in January this year, a fake AI-generated robocall purportedly from President Biden asked voters to skip a primary election (McMahon and Kleinman). This may be why Governor Newsom has been signing other bills focusing on misinformation and deep fakes- his blocking of the AI safety bill did not mean that he does not understand the risks that AI presents (Silva).

In conclusion, the EU’s AI Act may place them at a competitive disadvantage due to the regulations it enforces on AI developers, but it is clear that other countries have similar worries over the risks involved with AI. While others may be slower to regulate due to the concerns over hindering innovation, they will eventually need to and it could be beneficial to the EU to have a framework prepared and not be in a situation where the aforementioned risks do materialize and they can only legislate reactively. Moreover, as the US is at the forefront of this field, the AI Act may be the novel approach the EU requires to make their mark and, considering their position, it may hold less risk for them compared to countries with more established AI development.