The GPAI Code of Practice is a voluntary route to demonstrate AI Act compliance. Here is the official signatory list, the commitments in each chapter, and what happens when a provider does not sign.
The obligations for general-purpose AI model providers have applied since 2025. On 2 August 2026 the Commission's fine provision takes effect. Here is the distinction and a practical timeline.
The European Commission is proposing to delay some AI Act obligations for high-risk systems from August 2026 to December 2027, setting up a political battle between those calling for simplification and those warning of weaker safeguards.
Geoffrey Hinton, Yoshua Bengio, Steve Wozniak and hundreds of other prominent figures call for a ban on superintelligence development until it is controllable and has public backing.
SB 53 requires large frontier AI developers to publish safety protocols, report serious incidents and protect whistleblowers, helping California fill part of the federal regulatory void.
Anthropic will pay about $3,000 for each of roughly 500,000 books downloaded from pirate libraries. The settlement avoids a high-risk trial but does not establish that training AI on copyrighted works is illegal.
Starting today, China requires visible notices and technical markers for synthetic content published online. WeChat, Douyin and Weibo have already adapted their platforms to rules that also hold users and distributors responsible.
The U.S. will let Nvidia and AMD sell certain AI chips to China in exchange for 15% of the revenue. The deal turns export licenses into a source of government revenue and raises legal and national security questions.
As of August 2, providers of general-purpose models must meet new transparency and copyright obligations to market them in the EU. Models posing systemic risks face additional safety requirements.
The White House has unveiled its AI action plan, with more than 90 measures and three executive orders. It prioritizes data centers, tech exports and ideological criteria for federal procurement.
A federal judge has ruled that Meta’s use of copyrighted books to train Llama was fair use in the Kadrey case. But the decision warns that AI could still damage the publishing market without reproducing the original works.
Judge William Alsup hands down the first substantive ruling on fair use in AI training: training on books is legal, but Anthropic will face trial for using pirated copies.
This website uses cookies to improve the browsing experience. Cookie policy.
