According to a report published on Wednesday by Corporate Europe Observatory and LobbyControl, major technology companies, particularly in the US, have been lobbying the European Commission to water down the General-Purpose AI Code of Practice, which is intended to guide artificial intelligence providers on their compliance with the AI Act (see EUROPE 13629/7).
The conclusions of the investigation carried out are clear: “Big Tech enjoyed structural advantages from early on in the process and – playing its cards well – successfully lobbied for a much weaker Code than could have been”, say the authors of the report.
Google, Amazon, Microsoft, Meta and Apple have benefited from greater and easier access to the working groups and their chairs, as well as to many of the workshops organised by general AI model providers - the companies developing the large AI models that the Code is supposed to regulate.
Other multi-billion dollar AI companies, such as OpenAI, Anthropic and Hugging Face, have also found themselves included in these specific workshops - “each of which receive Big Tech funding”, the investigation denounces.
The drafting of the Code has been the subject of much negotiation and, recently, fierce criticism, particularly with regard to the details of the protection of fundamental rights and the copyright of publishers (see EUROPE 13613/25).
The new voluntary nature of certain measures, such as those aimed at preventing human rights violations, the risk of sexual abuse of children or privacy violations, has been widely denounced by many players in the sector, as well as by the MEPs involved in drafting the Code (see EUROPE 13608/6).
In early April, Reporters Without Borders (RSF) announced that it was leaving the discussions on the Code, highlighting the “absence of safeguards for the right to information”, the “the tech industry’s overwhelming influence in the process” and the “contempt” shown for issues of information protection, information issues and the risks of deepfakes and massive disinformation.
In addition, the investigation by the two NGO points to several cases of conflict of interest among the external consultants recruited by the Commission to help the AI Office draft the Code.
The consortium - Wavestone, Intellera and the Centre for European Policy Studies (CEPS) - all have ties with model providers, according to the authors of the report.
Wavestone received a ‘Microsoft Partner of the Year Award’ in late 2024 and works with Google Cloud and AWS. Intellera is a partner of technology suppliers such as Amazon, Google, IBM and NVIDIA. CEPS members include Apple, AWS, Google and Meta.
The NGOs fear that this offensive against the Code is a precursor to those that could be launched in the next wave of simplification of digital regulations, which could see the reopening of the regulation on the protection of personal data (see EUROPE 13627/10).
The final version of the Code is due to be presented on 2 May, but publication could be delayed until early summer.
See the report: https://aeur.eu/f/gma (Original version in French by Isalia Stieffatre)