Axios reported on August 28 that Europe's AI Office has not gone after any company covered under the AI Act for misconduct, though it can now request information or ask for access to models. Anthropic announced that future Claude models will generate text containing a watermark to comply with the Act, and OpenAI is publishing training data summaries.
Google and Meta both said in July they would adopt and develop transparency and watermarking tools. Euronews noted that most leading Western labs, with the notable exception of Meta, signed the EU's Code of Practice. Axios said the high-risk rules come later, in December 2027 and August 2028.
Source: Axios