THE European Union’s (EU) US$15 billion hammer against Big Tech is nearing completion, with regulation on artificial intelligence (AI) coming into force just as AI winners such as Microsoft and Nvidia are testing financial markets’ patience after big investments.
New rules aim to heighten transparency and manage risks posed by the likes of deepfakes, while also promoting investment and innovation in a region lacking tech champions. Breaches mean fines of as much as 7 per cent of annual revenue.
But there is a risk these new rules only end up entrenching the transatlantic tech divide.
Then there is the extra red tape on general-purpose AI such as OpenAI’s ChatGPT, which will require disclosure on training data and respect of copyright law. Clearly a good idea in principle, but there could be unintended consequences from the decision to create an extra sub-category for “systemically risky” high-powered models that could become the norm for all large models within a year.
BT in your inbox
Start and end each day with the latest news stories and analyses delivered straight to your inbox.
The extra requirements may disincentivise growth and keep small developers operating in a niche market, rather than encourage them to obtain a Brussels seal of approval.
The cause for concern is such that even the architect of the initial European Commission proposal behind the law says that its reach has ended up being too broad.
“The regulatory bar maybe has been set too high,” said Gabriele Mazzini, the architect and lead author of the proposal on the AI Act, which the European Parliament approved in March. “There may be companies in Europe that could just say there isn’t enough legal certainty in the AI Act to proceed.”
Mazzini fears this could end up entrenching big US tech companies that have the deep pockets and resources needed to cope with complexity and pay fines, while still investing in the technological frontier.
A karate enthusiast who spent time in the US startup world, Mazzini is a self-confessed technocrat who nonetheless admits that the overall Brussels rule-making machine may have blind spots. The incentives to keep adding regulation rather than subtracting it, and a lack of first-hand understanding of how startups operate, are partly responsible for the current problem.
Such effects already exist in other parts of the EU regulatory legacy, of course, from the sweeping General Data Protection Regulation, as well as new rules targeting “gatekeeper” firms such as Apple or Meta Platforms.
Fines in the billions have done little to dent Big Tech’s dominance. Tougher rules on anti-competitive behaviour and improper use of personal data are good to have, but come with trade-offs in the wider tech race.
High stakes
If the AI rules are especially worrisome to the likes of Mazzini, it is because the tech in its current form is more nascent than businesses such as social media or e-commerce – that and the high geopolitical and economic stakes make it all the more important to “regulate very thoughtfully”, he says.
Top officials such as Mario Draghi have sounded the alarm that the EU regulates while others innovate, and that the region’s economy needs the kind of productivity boost that AI could provide. Even pessimistic assessments that the tech could only end up adding 1 per cent of gross domestic product growth would still make it worth capturing at a time when Germany’s economy is shrinking.
It would be unfair to write off regulation before it is fully implemented. Bloomberg Intelligence’s Tamlin Bason reckons the risk of fines will not be enough to deter near-term AI investment. Furthermore, Roberto Viola, a top Commission official, told the Financial Times last month that the AI Act would help innovation and provide “ample space” for experimentation.
But the balance between harms and benefits can still be improved. One idea could be helping smaller firms bear the financial burden of compliance costs and provide tools to help anxious customers become “AI Act literate”, said Tobias Haar, general counsel at German startup Aleph Alpha.
Failing that, another idea could be to impose a sunset clause, or expiry date, on rules dealing with fast-moving tech. If Big Tech’s own AI expectations are having to be rewritten on the stock market, maybe regulators need their own pause for breath. BLOOMBERG
