欧州の新たなAI規制が施行へ、業界に影響

3 Min Read

Omar Havana via Getty Images

New rules governing the use of AI models are now in place in Europe.

The EU’s Artificial Intelligence Act was initially passed in 2024, and, starting on Aug. 2, became enforceable in the member states, with new rules designed to increase transparency for those using AI systems.

The act effectively makes the EU the world’s foremost regulator of AI, although that in itself is likely to pose challenges for how the rules are enforced, with no precedents to follow and the tech evolving so rapidly.

Couple these potential issues with the massive funds available to some of the world’s leading AI developers and vendors, and the U.S. administration’s sensitivity about what it perceives as unfair treatment of U.S. companies, and some potential flashpoints could lie ahead, with other jurisdictions likely to be monitoring how the EU fares in attempting to implement the regulations.

Nevertheless, the EU emphasized that user safety lies at the heart of the act. In a statement, it said: “As AI grows increasingly capable and integrated into everyday life, the AI Act helps ensure that AI is developed, deployed, and used safely, giving people and businesses across the EU greater confidence in the technology.”

Related:How AI is Reshaping Europe's Digital Sovereignty Debate

Among the tangible measures regarding transparency that users will notice are chatbots and other systems being required to disclose that they are interacting with AI; compulsory labeling of deepfakes; and AI-generated or altered content being required to carry machine-readable marks for easy detection.

Also introduced at this stage are the act’s rules for providers of general-purpose AI (GPAI) models, which are considered those that “can perform many different tasks and can be used in a wide range of tools and services, including AI agents.”

These rules extend to the most advanced GPAI models that pose potentially harmful systemic risks, including chemical, biological, radiological and nuclear incidents; and cyber offenses, including, as occurred recently, AI acting outside of human control.

A further requirement is that providers must give clear information on the content their models were trained.

Further down the line, meanwhile, will be the enforcement of strict obligations for the developers of AI used in “high risk” applications, such as health, safety and fundamental rights, which is currently scheduled for December next year.

Although some in the tech industry are concerned that enforcement of the act may slow development, more than 200 companies have already signed the EU’s Code of Practice on Transparency of AI-Generated Content -- a framework with guidelines for creating and labeling AI outputs -- including Anthropic, OpenAI, Meta, Microsoft and Google.

Related:Trump’s EO Furthers Model Exclusivity, Harming Cyber Defenders

Enforcement, meanwhile, will be mainly the responsibility of the recently established EU AI Office and national authorities.

“With the AI Act, we [have] established a clear, risk-based and durable framework for trustworthy AI -- one that gives innovators legal certainty while protecting the public interest,” Henna Virkkunen, the European Commission’s executive vice president for tech sovereignty, security and democracy, said in a statement. ““As enforcement begins, we are taking an important step toward AI that people and businesses can understand and trust.”

About the Author

Contributing Writer

Graham Hope has worked in automotive journalism in the U.K. for 26 years, including spells as editor of leading consumer news website and weekly Auto Express and respected buying guide CarBuyer.