The European Union has agreed on the details of the Artificial Intelligence Act, which represents a comprehensive set of rules targeting individuals who build and use artificial intelligence, with hopes of setting a historic precedent for other countries around the world.
After months of debate over how to regulate companies like OpenAI, lawmakers in the European Parliament, Council, and Commission, the three governmental branches of the European Union, met for more than 36 hours to finalize the new legislation. Under increasing pressure to reach an agreement before the start of the European Parliament election campaign in the new year.
“The European Union’s Artificial Intelligence Act is the first of its kind in the world,” said Ursula von der Leyen, President of the European Commission. “It provides a unique legal framework for the safe and responsible development of this technology, and for the protection of the safety and fundamental rights of individuals and businesses.”
Is this the first law for AI ?
While this law is not the first of its kind in the world, China’s new rules for generative artificial intelligence came into effect in August last year. However, the European Union’s law stands out as a comprehensive framework for the technology.
The European Union’s law prohibits the use of biometric systems that identify individuals based on sensitive characteristics, such as sexual orientation and race, and prohibits the random scanning of faces online. Lawmakers also agreed that biometric identification systems could be used in public places to investigate certain crimes.
What is included in this law?
The law also includes new transparency requirements for all general-purpose artificial intelligence models, including powerful models such as GPT-4. Companies that do not comply with the rules could be fined up to 7% of their global turnover. It is expected that the ban on some applications will come into force within six months, while transparency requirements will come into force within two years.
The law also includes measures to protect copyright in this field, requiring general-purpose models to be transparent about their energy use.
“The law sets rules for regulating large and powerful artificial intelligence models, which ensures that they do not pose systemic risks to the European Union,” said Dragos Tudorache, a member of the European Parliament.
Companies that do not comply with the rules could be fined up to 7% of their global turnover.
The ban on prohibited artificial intelligence will come into force within six months, while transparency requirements will come into force within 12 months, and the full set of rules is expected to come into force in about two years.
“Europe is leading the way in setting global standards, and we recognize the importance of our role as a standard-setter,” said Thierry Breton, European Commissioner, at a press conference.
Lawmakers have been negotiating for two years over the rules that have now been agreed upon, as artificial intelligence technology and key concerns have changed significantly.
There were concerns in April 2021 when the law was put into effect about opaque algorithms that influence decisions about jobs, asylum, and social benefits.
Examples were documented in 2022 showing the harmful effects of artificial intelligence, where algorithmic decisions led to the forced separation of Dutch families from their children, and students who studied online claimed that artificial intelligence systems discriminated against them based on the color of their skin.
In November 2022, OpenAI launched the ChatGPT chatbot, which revolutionized the debate.
In the final discussions of the law in Brussels, the main point of contention was whether law enforcement agencies could use facial recognition or other biometric measures to identify individuals either in real time or retrospectively.
