The European Parliament sets its position on the first major artificial intelligence regulation

On 14 June 2023, the European Parliament adopted its negotiating position on the proposed Artificial Intelligence Act. It was a decisive step rather than the final law: it enabled negotiations with the Council on an agreed text.

The timing mattered. The Commission’s original proposal dated from 2021, but the rapid spread of generative AI had introduced capabilities and risks that now dominated public debate.

A risk-based approach

The European framework started from a practical principle: not every use of AI should be treated in the same way. A content recommendation does not necessarily have the same impact as a system used in employment, education, critical infrastructure or access to essential services.

The approach distinguished levels of risk. Certain practices would be prohibited, high-risk uses would face stronger duties and other systems could be subject to transparency requirements or existing legislation. Context mattered as much as technology: purpose, data, affected people and the ability to appeal a decision all shape risk.

Generative and foundation models

By 2023, the debate had expanded to foundation models and generated content. Parliament proposed obligations concerning transparency, documentation and compliance with European rules, including the identification of generated or manipulated content in certain circumstances.

The discussion showed that responsibility could be shared across a chain: model developer, application provider, organisation supplying data and team using the output to make a decision. Using a third-party tool did not remove the need to understand its purpose, conditions and controls.

How businesses could prepare

Useful preparation did not need to wait for the final text. Organisations could inventory AI embedded in purchased software and internal experiments; assign an owner and purpose to each use; classify personal, confidential or protected data; review supplier terms; define human oversight; and record incidents.

Training was also part of control. Marketing, technology, HR and management encounter different data and risks. Teams need to recognise that fluent output may be wrong, that prompts can disclose information and that review must match the consequences of the use.

Innovation supported by trust

Regulation did not remove the opportunity to innovate. It sought conditions compatible with rights, safety and transparency. A company able to explain what it uses, why it uses it and which controls apply can turn trust into an advantage.

Sitelicon approaches AI as an enterprise system. We assess processes, data, permissions, architecture and metrics before automation. The June 2023 position opened a negotiation, but its lasting lesson was already clear: responsibility begins well before a regulation is fully applicable.

Editorial note: originally published in June 2023 and reviewed on 11 September 2026 to preserve the continuity of our editorial archive.