EU Leads the Way on Regulating AI

AI technology has dominated public discourse in recent months, raising anxieties over everything from job losses to human extinction. With a growing chorus of voices calling for legislation, the EU has its answer.

In 2021, critics deemed the EU’s move to introduce regulation on AI technology premature – but not two years on.

The EU AI Act has passed a successful plenary vote. It takes a risks-based approach, categorising AI applications into three risk categories: unacceptable risk, high-risk, and low-risk.

There was little in the way of opposition throughout the plenary debate, with broad acceptance of the need for rigorous regulation and a sense of pride that the EU was leading the way.

One point of contention that emerged, however, was the total ban on real-time surveillance. An unsuccessful amendment was tabled to allow flexibility for the use of this technology in cases of missing children or in efforts to prevent terrorist attacks. But the amendment raised significant concerns that safeguards would be abused.

Beyond technological imperfections, there is a risk posed by bad-actors who could abuse and misuse the technology, resulting in cases of innocent individuals incorrectly flagged as suspicious – an issue which raises serious concern over discrimination and racial profiling.

The EU’s AI Act is the first of its kind and could set the standards upon which other countries model their own legislation.

Rishi Sunak was pitching the UK as a possible global hub for AI regulation in a recent White House visit, but the reality is that Britain’s position as a global leader has invariably shrunk post-Brexit. With the country often perceived as increasingly untrustworthy and isolationist, the likelihood of the UK becoming a nucleus for this emerging technology appears to be another Brexit unicorn.

There have been some bold and worrying suggestions that AI technology could pose an existential risk for humanity, some of which have come from the creators of this technology themselves. But Romanian MEP Dragos Tudorache, co-rapporteur for the AI Act, has said he does not foresee a “Skynet” threat – in reference to the Terminator film franchise – but rather a series of “indirect induced threats and risks”. 

“When it comes to the already very divided and polarised society in which we live, the way AI can accelerate deepfakes or disinformation, if it’s not controlled and it’s not understood right, and managed right, then – again – it can make those rifts in our society so profound that at some point, democracy as we know it might actually implode,” he said.

In terms of the impact on the labour market, Tudorache believes that AI is going “to fundamentally effect some jobs in the workplace” and that “if we don’t prepare our workforce for that, if we don’t manage the effects of that, which are going to be quite profound for our society, for our people, then we might have a reaction in our society that we will not be able to control”.

Last month, BT announced significant job cuts, mostly in the UK, with around 10,000 positions being replaced by AI.

Deirdre Clune, an Irish Fine Gael MEP in the European People’s Party (EPP) and a lead negotiator on the AI Act for her party, said one of her biggest concerns was that “people don’t understand” the technology, citing deepfakes and misinformation being spread as a means of defrauding people.

Last month, an AI-generated image which suggested an explosion at the Pentagon caused a brief panic on the stock market. Clune called for greater “digital literacy” to combat misinformation. In terms of global cooperation, Clune suggested something similar to the “nuclear disarmament treaty, which countries sign up to” and an “international code of conduct” would be two possible mechanisms for greater global cooperation.

Developers will face a new slew of regulations and requirements under the Act, including an obligation to disclose the use of AI and to disclose what copyrighted material has been used to develop large language models such as ChatGPT.

The use of AI technology for social scoring, real-time biometric monitoring and emotion recognition systems will be banned outright, and high-risk applications will face further regulation. Breaches of the Act will carry a fine of 6% of global annual turnover. 

The legislation will not come into force until 2025 at the earliest – in the meantime, the European Commission is set to ask organisations who have signed up to its voluntary code of practice for combating misinformation to develop separate guidelines for dealing with AI generated misinformation.

The next era will be the era of technology. As global conferences are held in China, New York, and the UK, this technology will remain front and centre of political discourse while parliaments grapple with balancing regulation and innovation.

While the speed at which this technology is advancing may be alarming, there are significant benefits and opportunities to be had if harnessed correctly. For example, on climate change, advanced AI technology could monitor climate markers, identify trends and even create future models on the impact of those trends at a far-greater speed.

Balancing innovation and regulation comes down to the manner in which the balance is struck – and the EU appears steady in its convictions.