Europe Pioneering Global Efforts to Regulate AI

Global authorities are striving to harness artificial intelligence (AI), with the European Union (EU) set to surpass a significant legislative milestone this Wednesday.

The proposal — including contentious elements on facial recognition — is due to be voted on by the members of the European Parliament, moving it closer to being adopted into law.

The urgency to establish AI safeguards has increased in Brussels, as advances in chatbots such as ChatGPT underline both the potential benefits and risks associated with evolving technology.

Here’s a brief overview of the EU’s Artificial Intelligence Act:

HOW WILL THE REGULATIONS FUNCTION?

First introduced in 2021, the legislation will oversee any service or product that incorporates an AI system. It will categorize AI systems into four risk levels, ranging from minimal to unacceptable.

Applications deemed to pose higher risks, such as recruitment tools or technology aimed at children, will face stringent requirements, including higher transparency and accuracy in data use.

Breaches could incur fines of up to 30 million euros ($33 million), or 6% of a company’s global annual revenue. For tech giants like Google and Microsoft, the penalties could reach into the billions. The responsibility of enforcing these rules will fall to the EU’s 27 member states.

WHAT ARE THE CONCERNS?

A key aim of the EU is to protect against AI hazards to health and safety and to safeguard fundamental rights and values.

Therefore, certain AI applications, such as social scoring systems that evaluate people based on their behavior, are categorically prohibited.

Similarly, AI that exploits vulnerable individuals, such as children, or employs subliminal manipulation potentially causing harm, is also forbidden. For instance, an interactive talking toy prompting unsafe behavior would be unacceptable.

Tools for predictive policing, which use data to predict criminal activity, are also banned.

Parliamentarians have augmented the initial proposal by the European Commission, the EU’s executive branch, extending the ban on public facial recognition and biometric identification. However, an amendment by a center-right party has proposed exceptions for law enforcement, such as locating missing children, identifying suspects in serious crimes, or averting terrorist threats.

AI applications in areas like employment and education that could potentially influence a person’s life trajectory, must adhere to strict requirements including transparency to users and efforts to assess and reduce algorithmic bias risks.

The commission suggests that most AI systems, like video games or spam filters, fall into the low- or no-risk category.

WHAT’S THE DEAL WITH CHATGPT?

Initial provisions scarcely mentioned chatbots, except for a labeling requirement for user awareness. Provisions were later expanded to include general purpose AI like ChatGPT, which has rapidly gained popularity, thereby subjecting it to similar requirements as high-risk systems.

A key inclusion requires thorough documentation of any copyrighted material used to train AI systems to create text, images, videos, and music resembling human work.

This allows content creators to identify if their works have been used to train algorithms in systems like ChatGPT, and possibly seek remedies for copyright infringement.

WHY IS EU’s REGULATION CRUCIAL?

Although not a leading player in cutting-edge AI development, the European Union often sets the trend for global regulations, which frequently become global standards by default.

Given the vast consumer base of the EU’s single market, it’s often more feasible for companies to comply rather than create region-specific products.

But the intent is not solely regulatory. By establishing standard rules for AI, Brussels also aims to promote market growth by fostering confidence among users.

Other nations are expected to adapt and replicate the EU’s rules. Britain, since its exit from the EU in 2020, is positioning itself for AI leadership. Prime Minister Rishi Sunak plans to host an international summit on AI safety in the coming autumn.

The process before the regulations are fully operational could take several years. It will involve tripartite negotiations between member countries, the Parliament, and the European Commission, possibly leading to further amendments.

Approval is anticipated by the end of the year, followed by a grace period for companies and organizations to adjust, typically around two years.

In the interim, Europe and the U.S. are developing a voluntary code of conduct, which is promised to be drafted within weeks and could potentially be extended to other “like-minded countries.”

Frequently Asked Questions (FAQs) about AI Regulation

More about AI Regulation

• EU Artificial Intelligence Act
• European Parliament – AI and Robotics
• European Commission – Artificial Intelligence
• European Parliament News – AI Regulation
• Reuters – EU’s proposed AI rules
• TechCrunch – EU’s Artificial Intelligence Act
• Bloomberg – EU’s AI Regulation
• European Data Protection Supervisor – Opinion on the AI Act