EU strives to contain AI

Strict laws from the EU

This landmark law amendment is proposed by the European Commission to protect citizens from the dangers of technology.

Lawmakers from the European Parliament will vote on the proposal – including controversial amendments to facial recognition – when it is about to be passed. Brussels’s years-long effort to build protections against AI has become all the more urgent as the rapid advancements of chatbots like ChatGPT show the benefits emerging technology can bring, but it also comes with unpredictable risks.

This measure will govern any product or service that uses an artificial intelligence system. The Act will classify AI systems according to four levels of risk. Riskier applications, such as recruitment or children’s technology, will face tougher requirements, including transparency and accurate data usage.

Violations will result in fines of up to 30 million euros ($33 million) or 6% of the company’s annual global revenue, which in the case of tech companies like Google and Microsoft can amount to billions of dollars. . Enforcement of the rules will be up to the EU’s 27 member states.

One of the main objectives of the EU is to protect against any threat that AI may pose to human health and safety, and to protect fundamental rights and values. That means some uses of AI are completely undesirable, such as a “social scoring” system that rates people based on their behavior.

AI is also prohibited from exploiting vulnerable audiences, including children, or using potentially harmful subconscious manipulation, such as interactive talking toys that encourage dangerous behavior.

Lawmakers bolstered the original proposal from the European Commission (EC) by extending the ban on remote facial recognition and biometric identification in public.

But the ban faced a last-minute challenge after a center-right party added an amendment allowing law enforcement exceptions such as finding missing children, identifying suspects involved in serious crimes. address or prevent terrorist threats.

“We don’t want mass surveillance, we don’t want social scoring, we don’t want predictive policy in the EU, which needs to be completely stopped,” said Dragos Tudorache, a Romanian member of Parliament. Europe, co-leader on the AI ​​Act, said.

AI systems used in categories like employment and education, which will affect a person’s life, face tough requirements such as being transparent with users and taking steps to evaluate and reduce the risk of bias from algorithms.

The original laws barely addressed chatbots, mainly by requiring them to be labeled so that users knew they were interacting with the machine. Negotiators later added provisions to include general-purpose AI like ChatGPT after it exploded in popularity, making that technology subject to some of the same requirements as systems that have high risk.

Promoting AI-controlled communities

The sheer size of the EU single market, with 450 million consumers, makes it easier for companies to comply than to develop different products for different regions, experts say. By setting general rules for AI, Brussels is also trying to develop the market by creating trust among users.

“The fact that this is an enforceable regulation and that companies will be held liable is very important because countries like the US, Singapore and the UK are merely providing guidance and recommendations,” he said. Kris Shrishak, a technologist and senior fellow at the Irish Civil Liberties Council.

Other countries are also trying to catch up with the EU. The UK is vying for the AI ​​leadership position as Prime Minister Rishi Sunak plans to host a world summit on AI safety this fall.

Prime Minister Sunak said at a technology conference this week: “I want the UK to be not only the intellectual home but also the geographical home of global AI safety regulation. The UK summit will bring together people from academia, business and government from around the world to work on “a multilateral framework”.

It may be many years before the rules come into full force. The vote will be followed by tripartite talks involving EU member states, the European Parliament and the EC, which could face more changes when trying to agree on wording. But final approval is expected later this year, after which it is time for companies and organizations to adapt, usually around two years.