The G7’s main tool for corporate AI disclosure is entering its second reporting cycle. The OECD launched version 2.0 of the Hiroshima AI Process Reporting Framework on May 28, 2026, at a Tech7 event in Paris on the margins of the G7 Digital and Tech Ministerial under the French presidency. It encouraged organizations to submit reports under the revised framework by September 30, 2026, to be included in its next planned analytical review.
The framework is voluntary. The OECD describes it as “a common, voluntary structure for organisations to disclose their AI governance and risk management practices.” For companies that build or deploy AI, it is becoming a shared reference point that regulators, investors and customers can read across jurisdictions. More than 50 organizations have pledged to complete a report under version 2.0, according to the OECD, including Amazon, Google, Microsoft, Mistral AI and OpenAI.
What changed in version 2.0
The OECD says the new version is simpler and more structured, with the aim of opening it to small and medium-sized enterprises as well as large developers. Questions are now tailored to an organization’s role in the AI value chain: model developer, application developer or deployer. The framework covers newer technology, including agentic AI.
Version 2.0 links to the OECD.AI Catalogue of Tools and Metrics for Trustworthy AI, so organizations can cite tools they already use or add new ones. It is aligned with the Hiroshima AI Process Code of Conduct and other risk management approaches, which the OECD presents as a way to support interoperability. The framework was refined through a pilot with organizations from seven countries. Submissions are accepted on a rolling basis, and organizations are invited to update their reports each year.
Where it sits in the G7 agenda
The Hiroshima Code of Conduct dates from Japan’s 2023 G7 presidency. The reporting framework followed in February 2025, and an OECD analytical report presented preliminary insights from submissions by 20 organizations.
France has kept the work on its agenda. At their meeting in Paris on May 29, G7 digital ministers, chaired by Anne Le Hénanff, France’s minister delegate for AI and digital, agreed to continue work under the Hiroshima AI Process. They also adopted, for the first time, a common vision on the different degrees of openness of AI systems, intended to give companies and researchers shared reference points. With the OECD, the presidency launched a G7 SME AI Readiness Tool to help smaller firms assess their readiness and needs.
What to watch
The next marker is the OECD’s analytical review of reports filed under version 2.0. The OECD page does not give a publication date. For public affairs teams, the practical questions are whether their organization has filed, how its disclosures compare with those of peers, and whether governments begin to cite the framework in national rules or procurement. The French presidency’s G7 agenda runs to the end of the year, and the presidency then passes to the next host.















