The European Union (EU) unveiled fresh guidelines under its AI Act focusing on prohibited uses of the technology, as it pushes forward with the regulatory framework that has attracted opposition from industry heavyweights.

The AI Act, first unveiled in 2023, categorises AI technologies into minimal, transparency, high, or unacceptable risk levels, moving to ban systems that pose a threat to safety and fundamental rights. The latest measures, released by the European Commission on Tuesday (4 February), provide detailed guidance on how the rules will apply.

Under the new guidelines, which came into effect from 2 February, AI systems that carry out “manipulative, exploitative, social control or surveillance practices” are barred from the EU. This includes AI practices such as predictive social scoring, mass biometric surveillance and scraping internet sources or CCTV footage to build facial recognition databases.

The legislation also mandates transparency requirements for companies developing “high-risk” AI systems, which serve public service functions including healthcare, credit-scoring, and migration. Developers of more advanced AI models face even stricter obligations including mandatory human oversight of AI systems and risk assessments.

National regulators will oversee the enforcement of the act, with non-compliant companies facing hefty financial penalties of up to €35 million or higher with respect to their global annual revenue. Companies engaging in prohibited AI practices could also be hit with a ban from operating in the EU.

However, the framework exempts select areas from its scope including national security tasks, third-country authorities or international organisations using AI for law enforcement, and pre-market AI research and development.

Industry pushback
Several tech giants including Meta Platforms have raised concerns over the stringent regulations, arguing they could crush AI innovation and investment. In response to the latest announcement, head of global affairs at Meta, Joel Kaplan, announced the company will not join the EU’s AI Code of Practice, calling it “unworkable” and claiming it imposes unnecessary burdens on open-source AI models.

Meanwhile, US President Donald Trump has warned the EU against targeting US tech companies, while hinting at possible retaliatory measures targeting businesses in the bloc. “We have some very big complaints with the EU,” he said last month.

Despite the criticism, an EU official told Financial Times the regulatory body’s stance on AI remains firm. “What we can do is ensure that it is as innovation-friendly as possible, and that’s what we’re doing right now,” he added.