The European Parliament has approved the world's first comprehensive framework for constraining the risks of artificial intelligence (AI).

The sector has seen explosive growth - driving huge profits but also stoking fears about bias, privacy and even the future of humanity.

The AI Act works by classifying products according to risk and adjusting scrutiny accordingly.

The law's creators said it would make the tech more "human-centric."

"The AI act is not the end of the journey but the starting point for new governance built around technology," MEP Dragos Tudorache added.

It also places the EU at the forefront of global attempts to address the dangers associated with AI.

China already has introduced a patchwork of AI laws. In October 2023, US President Joe Biden announced an executive order requiring AI developers to share data with the government.

But the EU has now gone further.

"The adoption of the AI Act marks the beginning of a new AI era and its importance cannot be overstated," said Enza Iannopollo, principal analyst at Forrester.

"The EU AI Act is the world's first and only set of binding requirements to mitigate AI risks," she added.

She said it would make the EU the "de facto" global standard for trustworthy AI, leaving every other region, including the UK, to "play catch-up."

In November 2023, the UK hosted an AI safety summit but is not planning legislation along the lines of the AI Act.

How the AI Act will work

The main idea of the law is to regulate AI based on its capacity to cause harm to society; the higher the risk, the stricter the rules.

AI applications that pose a "clear risk to fundamental rights" will be banned, for example some of those that involve the processing of biometric data.

AI systems considered "high-risk", such as those used in critical infrastructure, education, healthcare, law enforcement, border management or elections, will have to comply with strict requirements.

Low-risk services, such as spam filters, will face the lightest regulation - the EU expects most services to fall into this category.

The Act also creates provisions to tackle risks posed by the systems underpinning generative AI tools and chatbots such as OpenAI's ChatGPT.

These would require producers of some so-called general-purpose AI systems that can be harnessed for a range of tasks, to be transparent about the material used to train their models and to comply with EU copyright law.

Mr Turodache told reporters ahead of the vote that copyright provisions had been one of the "heaviest lobbied" parts of the bill.

OpenAI, Stability AI and graphics chip giant Nvidia are among a handful of AI firms facing lawsuits over their use of data to train generative models.

Some artists, writers and musicians have argued the process of "scraping" huge volumes of data, including potentially their own works, from virtually all corners of the internet violates copyright laws.

The Act still has to pass several more steps before it formally becomes law.

Lawyer-linguists, whose job is to check and translate laws, will scour its text and the European Council - composed of representatives of EU member states - will also need to endorse it, though that is expected to be a formality.

In the meantime, businesses will be working out how to comply with the legislation.

Kirsten Rulf - a former advisor to the German government, and now a partner at Boston Consulting Group - says more than 300 firms have been in touch with her company so far.

"They want to know how to scale the tech, and get value from AI," she told the BBC.

"Businesses need and want the legal certainty." (BBC)

• AI symbol

Update news