Microsoft announced Thursday that it has signed the EU’s voluntary Code of Practice for general-purpose AI model providers.
The code lays out steps that companies can take in order for the European Commission to consider them compliant with the AI Act, which regulates artificial intelligence in general.
Companies commit to assessing and tackling certain risks, for example whether an AI model discriminates against groups of people or could escape from human oversight.
France’s Mistral, Germany’s Aleph Alpha, as well as Anthropic and Google have said they will sign the code. Meta, on the other hand said it will not, arguing the code introduces “a number of legal uncertainties”.
Microsoft walked a middle path in its announcement Thursday.
The company said it wanted to build further trust in its AI models and demonstrate compliance with EU law “while recognizing that the AI Act is a complex regulation that would benefit from simplification”.
It called on European policymakers to use the upcoming digital omnibus package “to refine and improve the AI Act and Code of Practice”.
(vc)