A third version of the code of conduct for ‘general-purpose’ AI systems has been published. The softer tone of the text suggests a different approach to regulation.
The European Union announces in a press release that a third version of the AI Code of Practice has been published. This document lists the key rules and obligations for ‘general-purpose’ AI systems, such as ChatGPT, based on the AI Act. The text is written by a committee of independent experts.
read also
First AI Act rules in effect: Overview of prohibited AI practices and obligations
The code of conduct is divided into three sections: transparency, copyright, and risk analysis. The third section only applies to the most powerful AI models, while certain open-source models are exempted from the transparency obligation. This is not yet the final version: after a feedback round until March 30, a new text will be published in May.
Softer Tone
What stands out amidst the formal and legal language is that throughout the entire code of conduct, a softer tone is used compared to the second versions. Those focused mainly on the obligations of AI companies and sounded much more imperative. Now, there is more pragmatic and solution-oriented thinking, with a focus on collaboration and future-proofing.
The shift is particularly evident in the copyright section, TechCrunch notes. Words like ‘immediately’ and ‘quickly’ are replaced with terms such as ‘to the best of ability’, ‘reasonable’, and ‘proportionate’ when it comes to measures AI companies must take to prevent their models from infringing copyrights. The text even suggests that companies may ignore complaints if they are ‘excessive’, ‘unfounded’, and/or ‘repetitive’.
Pressure from Inside and Outside
What will and won’t be in the final text remains to be seen, but the renewed code of conduct suggests a different approach from the European Union. The new legislature seems to want to take a softer stance towards Big Tech companies than the previous one, which kept large tech companies on a tight leash with strict regulations. The United States is putting pressure on the European Union to soften regulations. The European Commission had previously rejected the liability law of the AI Act.
read also
Europe’s digital identity crisis: “Regulation is not us-against-them”
But within the European tech industry as well, support for the AI Act and regulation is waning. This became clear during MWC: the call that Europe is lagging behind in the AI race echoed through the exhibition halls. Among others, Mistral CEO Arthur Mensch raised the alarm during the fair. Claudia Nemat, board member of Deutsche Telekom, poetically expressed the sentiment: “In Europe, we invented the seatbelt, while the car doesn’t even exist yet”.