The EU AI Act ‘reflects the reality of what’s actually happening’

Following the landmark passing of the EU AI Act, SiliconRepublic.com spoke to William Fry’s Barry Scannell to take a closer look at what it all means.

The AI goldrush is far from slowing down right now as Big Tech continues to pour investment into the explosive trend.

However, while companies continue to capitalise on AI – particularly generative AI – with new tools and products, the EU passed its landmark AI Act last month, which seeks to rein in high-risk tech and prohibit certain uses like social scoring.

The Act was reportedly delayed earlier this year due to the rise of foundational models such as OpenAI’s ChatGPT, with lawmakers eager to include rules around this particular tech.

To delve further into the new Act and explore what the ramifications of these new rules will be, SiliconRepublic.com spoke to Barry Scannell, a consultant in William Fry’s technology department.

He said that it was important that AI regulation started somewhere and that it’s a good thing that the EU is taking the lead.

“I’ve been using the example of the David Attenborough documentary of the penguins standing on the ice and waiting for the first one to jump in,” he said.

“That doesn’t apply just to the EU and who’s regulating it, it also applies to companies and deciding how they’re going to go about implementing their own frameworks and governance as well.”

Several companies at the forefront of this AI race – OpenAI, Google and Microsoft to name but a few – have already been quite vocal about AI regulation.

In a bid to stay on regulators’ good sides both in Europe and the US, several AI players have committed to “robust” mechanisms such as watermarking. A number of the companies have also come together to form an industry body to promote responsible AI research and development.

For other companies, Scannell said it’s important to use the EU AI Act as a good place to start when ensuring compliance, even though it’s not yet in force.

“It’s still a really good framework in which to base your AI governance on because it catches things like risk management, data governance and things like that.”

Real-time biometrics

One of the biggest areas reined in thanks to the EU Act is biometric identification and surveillance. In particular, real-time remote biometric identification is prohibited.

An example of this would be people walking down a public street where there are CCTV cameras. Real-time remote biometric identification would mean identifying everyone using biometrics in real time.

In recent years, controversial facial recognition company Clearview AI got into hot water for gathering data without consent, leading to a large fine in the UK and a ban in Canada.

In Ireland, there was also debate around the use of facial recognition within the Gardaí. While the EU AI Act outlaws the use of real-time facial recognition technology, post-remote biometric identification will be permitted in very limited circumstances.

“Post-remote biometric identification would be that you can go back after the fact, get CCTV footage and then apply facial recognition to it, but that would only be permitted in circumstances basically where you have something like a warrant from a judge, so it’ll be almost on the level of a search warrant for someone’s house to help protect people’s fundamental rights.”

With the technology moving so fast, could there be a concern that the AI Act won’t be able to keep up? Scannell is optimistic that the current definitions within the legislation effectively capture the new iterations.

“The EU AI Act reflects the reality of what’s actually happening,” he said. “The next leap forward in AI probably won’t come from the GPT transformer-type technology because it’s a brute force type of technology. It’s based on consuming all the data and that’s not very efficient from a computing point of view, from an energy point of view and from many different points of view.”

He added that while there’s still loads of work to do in the AI world, the EU is doing is definitely going in the right direction.

Earlier this week, the latest episode of For Tech’s Sake was released, featuring Ireland’s AI ambassador Dr Patricia Scanlon, who also spoke about the EU AI Act, growth of generative AI and the importance of ethical AI.

10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.