On Friday, the European Commission published guidelines for AI models with systemic risks. These are intended to help companies comply with the AI Act, which will come into force for these models at the beginning of August.
The AI Act became law last year and will take effect for models with systemic risks as of August 2. Companies such as Google, OpenAI, Meta, Anthropic, and Mistral have until August 2 next year to comply with the legislation.
This move aims to address criticism from companies about the regulatory burden. At the same time, the Commission aims to provide greater clarity to companies that risk fines ranging from €7.5 million to €35 million, or 1.5 to 7 percent of their global annual turnover.
Stricter rules for powerful AI models
The new guidelines focus on AI models that the Commission has identified as posing systemic risks. These are models with highly advanced computing capabilities that can have a significant impact on public health, safety, fundamental rights, or society.
On August 2, these companies will be required to evaluate their models, assess, and mitigate risks. They must also conduct adversarial testing, report serious incidents to the Commission, and provide adequate cybersecurity protection against theft and misuse.
Transparency requirements for foundation models
General-purpose AI models or foundation models will be subject to transparency requirements. These companies must prepare technical documentation, establish copyright policies, and provide detailed summaries of the content used to train algorithms.
“With today’s guidelines, the Commission supports the smooth and effective application of the AI Act,” said EU Tech Commissioner Henna Virkkunen.
The guidelines come despite persistent criticism from the business community. CEOs from Siemens and SAP recently renewed their call for a review of European AI legislation. They argue that the current rules hinder innovation and that Europe is losing ground to other regions as a result.
The Commission appears to be sticking to its planned timeline for now. With these guidelines, Brussels aims to provide companies with the necessary tools to operate in accordance with the new rules without compromising Europe’s innovative strength.