Europe concedes more ground to US, Big Tech on AI rules

Europe’s push for digital independence faces a setback as officials ease artificial intelligence rules while one of the continent’s most successful AI companies hands over its infrastructure to an American tech giant.

The deal is tentative and needs official approval before it’s final. It happened after the talks dragged too long between the country’s representative and parliament members, according to Reuters.

The most significant change postpones requirements for high-risk AI systems covering biometric identification, critical infrastructure, and law enforcement. Originally scheduled to start this year, these rules will now kick in at the end of 2027.

Some industries will be exempted once the law comes into action. This includes machine manufacturers. Equipment that existing industry regulations already cover will stay outside the AI Act’s reach. The European Commission made this adjustment after companies complained about duplicate rules and extra paperwork.

European businesses have spent recent years saying new laws slow down innovation. Therefore, the deal is being worked on to give space to EU firms to up their game against US rivals. However, it has also attracted criticism over how heavily policymakers are influenced by big tech companies.

While some rules get relaxed, others get stricter. The EU will ban AI tools that create sexually explicit images of people without their permission. AI-created content will also need visible watermarks or labels starting in December of this year.

Kim van Sparrentak, a Dutch member of the European Parliament, said the ban on explicit deepfakes aims mainly to shield women and children from harmful uses of generative AI technology.

German translation leader partners with Amazon

These regulatory shifts come at an awkward time for Europe’s AI sector. DeepL, a translation company based in Cologne, Germany, recently announced it would work with Amazon Web Services. The move has industry watchers worried about Europe losing its edge in machine translation.

DeepL has built a strong reputation by consistently beating Google Translate in accuracy tests. Governments, courts, and half the companies on Fortune’s list of America’s 500 top earners use its services. The company brought in $185.2 million last year. Last month, it rolled out a live voice-to-voice translation feature.

DeepL told paying customers it would stop handling data only on its own servers. The company said it needed Amazon Web Services to grow internationally.

Concerns over data control and American laws

Jörg Weishaupt runs Malogica Group, a software company in Madeira, Portugal. He had used DeepL for years but decided to cancel after the Amazon announcement. He told The Guardian he no longer feels safe uploading contracts or internal strategy documents. “These are confidential documents, and I want to know where they end up,” he said.

DeepL responded that Amazon would not see or use customer data. A company representative said customer information gets encrypted and is not used to train AI models.

Weishaupt pointed to two American laws. The 2001 Patriot Act and 2018 Cloud Act, that let the US government ask cloud providers for information.

Last July, a Microsoft legal director told a French hearing the company cannot promise EU customers their data stays protected if the Trump administration asks for access to information on Microsoft servers.

DeepL offers a data residency option promising information stays in Europe, but some doubt whether such promises hold up.

If you're reading this, you’re already ahead. Stay there with our newsletter.