The public is locked out

“This is private data”, “disclosing information would affect commercial interests”, or “those meetings took place, but no minutes were recorded”. These are just some of the responses European governments gave when asked – via Freedom of Information requests – to disclose their meetings with major technology companies regarding Artificial Intelligence regulation. While regulatory frameworks are advancing across Europe, secrecy remains the norm when it comes to the decision-making process. Many European governments are withholding critical details about who they are meeting with and what is being discussed. However, these conversations between governments and tech companies could directly shape AI policies affecting millions.

Regulating AI is one of the most urgent and complex challenges policymakers face today. Across Europe and the world, governments are struggling to balance rapid technological innovation with ethical, social, legal and economic concerns. The European Union has developed the Artificial Intelligence Act (Regulation (EU) 2024/1689), a groundbreaking regulatory framework poised to set a global precedent. At the same time, many European governments are independently shaping their national approaches to deal with AI. To do so, they hold regular meetings with representatives from major technology companies. 

However, while meetings between EU institutions and corporate representatives are relatively accessible, many national governments operate under a veil of secrecy. Which governments are meeting with tech companies to discuss AI regulation? What is being discussed, which proposals and promises are made? And how do they shape national policies? This investigation used access to information laws to seek answers. 

Over the course of a year, we encountered delays, refusals, and exemptions – many citing commercial interests or privacy concerns – blocking the release of crucial information. The findings signal a lack of transparency, raising urgent questions about who is truly influencing the future of AI regulation in Europe.

What we know: Meetings with Google or Facebook and discussion of AI models

The current data shows that meetings between companies and Ministries on AI regulation are happening across multiple countries. However, the full extent and substance of these conversations remain difficult to understand as governments did not disclose all relevant information. A few examples:

United Kingdom: On the 1st of March 2023, the Department for Business and Trade in the UK discussed the differences between EU and UK AI regulations in a meeting with Microsoft, noting that “the EU is considering excluding AI from broad sectors of critical national infrastructure, whereas the UK is taking a case-by-case approach.” However, all other details from the discussion were redacted.

Denmark: The Ministry for Digital Government and Gender equality met with a representative from Facebook in January 2023, but no minutes were recorded. There are, however, meeting notes from February 9, 2023, detailing a discussion with Microsoft Denmark. The agenda included discussions on the minister’s and the government’s provisional visions and priorities for digitalisation. On March 8, 2023, there was another dialogue with Microsoft regarding the Danish AI sandbox (an initiative by the Danish Data Protection Authority that provides companies and authorities with free access to guidance on legal frameworks), during which the Department presented on AI regulation. No agenda or list of participants were recorded.

Germany: The Federal Ministry for Economic Affairs and Climate Action shared documents related to the AI Roundtable held on April 20, 2023. According to the Ministry, the goal of the meeting was to “raise awareness” among company management about the importance of large AI models for future competitiveness. However, the list of attendees was heavily redacted, naming only one company. The Ministry also noted that on March 24, 2023, a discussion on AI took place between representatives from the Ministry and Alphabet during a business trip to San Francisco. No documents related to this meeting were made available.

Ireland: The Department of Enterprise, Trade and Employment provided a more comprehensive response to the requests. It shared a schedule of records  including some minutes from meetings and correspondence with Google, Facebook, and Microsoft. While some information was redacted due to concerns over the commercial interests of the companies, the provided documents revealed discussions between the Ministry and Microsoft on activities aimed at building public trust and awareness of AI. Additionally, they provided heavily redacted emails in which Google representatives shared their views on several AI policy regulation developments in Europe with the Ministry. Among the redacted parts were Google’s views themselves.

What we can’t know: Commercial interests, fees or no notes taken

While most ministries responded to information requests, many provided incomplete or heavily redacted answers, leaving citizens in the dark about the nature of their interactions with technology companies. Refusals to disclose information were justified using the following exemptions:

Personal Information: Some Ministries like the Department for Business and Trade in the UK confirmed they held the information requested but released a heavily redacted document in which only the companies that attended the meeting were visible but not the participants from the side of the Ministry. They argued that releasing names of all Ministry attendees might infringe their right to privacy.

Commercial Interests: Some Departments in the UK and Ireland argued that full disclosure of the information would be contrary to the legitimate expectations of confidentiality of companies and the release would be likely to damage the commercial interests of the companies.

Exceeds cost limit: The Department for Science, Innovation and Technology in the UK argued that compiling the information would exceed the cost limit. The Bundeskanzleramt in Germany argued that compiling the information would require a great deal of effort and if the request were approved, an unspecified fee would have to be paid.

Internal notes with no relevant public interest: Spain’s Ministry of Economic Affairs and Digital Transformation argued that any minutes of potential meetings between the Ministry and tech companies would qualify as internal notes, and therefore not be subject to FOIA requests.

A meeting took place but there were no minutes: Some Ministries acknowledged the existence of meetings between the Ministry and tech companies but stated that no minutes were recorded, such as the Ministry of Digital Government and Gender Equality in Denmark.

Why transparency on AI regulation matters

The EU has been actively shaping AI regulation over the past years. Its approach includes multiple regulatory instruments, and at the core of this framework is the AI Act, the world’s first comprehensive AI regulation, adopted in 2024. It follows a risk-based approach, classifying AI systems by risk level and assigning specific obligations accordingly. Big tech companies have been working to dilute the EU AI Act by lobbying for weaker regulations and seeking exemptions for their most advanced AI systems. Their efforts include pushing for vague definitions, minimising compliance requirements, and emphasizing voluntary codes of conduct over binding legal obligations. These strategies aim to reduce regulatory burdens, potentially at the expense of public accountability and fundamental rights. Proponents of ethical AI have long argued to limit certain systems due to its recorded detrimental effects such as discrimination and privacy intrusion.

Freedom of Information Acts, also known as Access to Information Laws or Transparency Acts, empower citizens to access information held by public authorities. These laws outline government responsibilities, such as the proactive disclosure of data and, more importantly, the right for individuals to request specific information. They also set deadlines for authorities to respond to requests and outline the exemptions that can be argued by authorities to deny the release of information.

However, obtaining information on AI regulation from national governments within the EU has been a challenging and uneven experience. Across Europe, significant disparities exist in how governments handle similar requests, revealing that citizens’ rights to access information and hold their leaders accountable differ depending on where they live. While some governments simply declared a lack of communication with AI lobbyists, others provided limited insight into the regulatory process. Despite the adoption of the AI Act, AI regulation in Europe remains an evolving process shaped by multiple stakeholders with often conflicting interests. Therefore, transparency is crucial in the AI regulation process, as it ensures accountability and allows citizens to understand how decisions that affect their lives are made.

→ see all documents from all countries