Legislators have provisionally agreed to sweeping new laws that will regulate the use of artificial intelligence (AI) in Europe, including controls around the use of copyrighted music.

The deal between policy makers from the European Union Parliament, Council and European Commission on the EU’s Artificial Intelligence Act was reached late on Friday night in Brussels local time following months of negotiations and amid fierce lobbying from the music and tech industries.   

The draft legislation is the world’s first comprehensive set of laws regulating the use of AI and places a number of legal obligations on technology companies and AI developers operating in Europe, including those working in the creative sector and music business.   

The precise technical details of those measures are still being finalized by EU policy makers, but earlier versions of the bill decreed that companies using generative AI or foundation AI models like OpenAI’s ChatGPT or Anthropic’s Claude 2 would be required to provide summaries of any copyrighted works, including music, that they use to train their systems. 

The AI Act will also force developers to clearly identify content that is created by AI, as opposed to human works before they are placed in the market. In addition, tech companies will have to ensure that their systems are designed in such a way that prevents them from generating illegal content. 

Large tech companies who break the rules – which govern all applications of AI inside the 27 member block of EU countries, including so-called “high risk” uses — will face fines of up to €35 million or 7% of global annual turnover. Start-up businesses or smaller tech operations will receive proportionate financial punishments, said the European Commission.   

Governance will be carried out by national authorities, while a new European AI Office will be created to supervise the enforcement of the new rules on general purpose AI models. The pact was brokered after three days of talks between the European Union Parliament, Council and European Commission in the fifth and scheduled-to-be-final session of so-called ‘trilogue’ talks.

European legislators first proposed introducing regulation of artificial intelligence in 2021, although it was last year’s launch of ChatGPT — followed by the high-profile release of “Heart on My Sleeve,” a track that featured AI-powered imitations of vocals by Drake and The Weeknd — that alerted music executives to the technology’s potential impact on the record business. In response, lobbyists convinced lawmakers to add transparency provisions around the use of music and training data by developers.   

While the exact nature of those conditions are still to be confirmed, the European Commission said on Friday (Dec. 8) that the legislation introduces dedicated rules for generative AI models “that will ensure transparency along the value chain.”

President of the European Commission Ursula von der Leyen called the agreement “a historic moment” that “will make a substantial contribution to the development of global rules and principles for human-centric AI.” 

Responding to the announcement, Tobias Holzmüller, CEO of German collecting society GEMA, said the successful conclusion of talks was a welcome “step in the right direction” but cautioned that the act’s rules and provisions “need to be sharpened further on a technical level.”   

“The outcome must be a clearly formulated transparency regime that obliges AI providers to submit detailed evidence on the contents they used to train their systems,” said Holzmüller.  

Representatives of the technology industry, which had lobbied to weaken the AI Act’s transparency provisions, criticized the deal and warned that it was likely to put European AI developers at a competitive disadvantage.  

Daniel Friedlaender, Senior Vice President of the Computer and Communications Industry Association (CCIA), which counts Alphabet, Apple, Amazon and Meta among its members, said in a statement that “crucial details” of the AI act are still missing “with potentially disastrous consequences for the European economy.”  

“The final AI Act lacks the vision and ambition that European tech start-ups and businesses are displaying right now,” added CCIA Europe’s Policy Manager, Boniface de Champris. He warned that, if passed, the legislation might “end up chasing away the European champions that the EU so desperately wants to empower.”  

Now that an political understanding has been reached on the AI Act legislators will spend the coming weeks finalizing the exact technical details of the regulation and translating its terms for the 27 EU member countries.   

The final text then needs to be approved by the European Council and Parliament, with a decisive vote not excepted to take place until early next year, possibly as late as March. If passed, the act will be applicable two years after its entry into force, except for some specific provisions: bans will apply after six months while the rules on generative AI models will begin after 12 months. 

The first-of-its-kind legislation will apply to any company that operates in the European Union, regardless of where they are based. It comes as other countries, including the United States, China and the United Kingdom, explore their paths to policing the rapidly evolving AI sector.  

In a statement, IFPI, the global recorded-music trade organization, said the “groundbreaking” act provides “a constructive and encouraging framework” for regulation of the nascent technology.    

“AI offers creators both opportunities and risks,” said an IFPI spokesperson, “and we believe there is a path to a mutually successful outcome for both the creative and technology communities.”