AI Regulation Tracker / Standards and frameworks

OECD Plans Streamlined Hiroshima AI Reporting Framework v2

Q: Is the Hiroshima reporting framework mandatory?

No. The OECD describes it as a common, voluntary structure for organisations to disclose their AI governance and risk management practices. Nobody is required to file, and there is no penalty for organisations that choose not to. It is a transparency template tied to the G7's voluntary Code of Conduct, not a law.

The OECD is streamlining the Hiroshima AI Process (HAIP) Reporting Framework, the voluntary template that lets organisations disclose how they govern and manage the risks of advanced AI systems. In a March 2026 update, the OECD said a simplified v2.0 "will be available shortly." This is a voluntary transparency tool, not a binding rule, and it carries no penalty for organisations that choose not to file.

By Anthony Guerriero · The Leveraged Years Regulatory Research · Published May 28, 2026 · Last updated May 28, 2026 · Standard, voluntary framework

The short version

The OECD runs a reporting framework tied to the G7's Hiroshima AI Process Code of Conduct. In the OECD's own words, it gives organisations "a common, voluntary structure for organisations to disclose their AI governance and risk management practices." It is a disclosure template, not a law. Nobody is compelled to file, and there is no penalty attached to sitting it out. Launched in February 2025 as the first international, voluntary tool of its kind, it drew submissions from 20 organisations in an early round. The OECD said in a March 2026 update that it was streamlining the framework and that a simpler v2.0 "will be available shortly," with a pilot open at OECD.AI/Transparency. Some secondary reporting later pointed to a v2.0 around a G7 ministerial in May 2026, but the OECD press release was not reachable at the time of writing, so that timing is unverified. Who should care: organisations that build or deploy advanced AI, and the lawyers, CPAs, and consultants who advise them. Even though filing is optional, a public HAIP report is becoming a reference point that customers, procurement teams, and other regulators can read. Primary source: oecd.ai/en/hiroshima.

Regulatory Briefing

Instrument: Hiroshima AI Process (HAIP) Reporting Framework, forthcoming streamlined v2.0, under the G7 Hiroshima AI Process International Code of Conduct for Organisations Developing Advanced AI Systems
Authority: OECD, administering the reporting framework for the G7 Hiroshima AI Process
Jurisdiction: International (G7 process); open to organisations across jurisdictions
Status: Live and voluntary. Original framework launched February 2025; in a March 2026 update the OECD said a streamlined v2.0 "will be available shortly," with a pilot at OECD.AI/Transparency
Bindingness: Non-Binding Framework. SIGNAL, not action. It is a voluntary disclosure template. It does not require reporting, and it imposes no penalty for not filing.
Relevant date: February 2025 (original launch); streamlined v2.0 announced March 2026 as forthcoming
Primary Source: OECD, "Hiroshima AI Process Reporting Framework," oecd.ai/en/hiroshima
Workflow Impact: Firms weighing whether to publish a HAIP report now have a shorter, simpler template. The disclosure itself is a governance artifact your counsel and audit function will want to review before it goes out.
Operational Next Step: If you develop or deploy advanced AI, read the framework and a few published reports before deciding whether a voluntary filing helps or hurts your position.
Last Verified: July 23, 2026

Start with what the framework is, because the name makes it sound heavier than it is. The Hiroshima AI Process is the G7's workstream on advanced AI. It produced a Code of Conduct for organisations building the most capable systems, and the code is voluntary. The reporting framework sits one layer below that. It is a standard template that lets a company describe, in a comparable format, how it identifies and manages AI risk, how it handles transparency, how it governs the work, and what it does on content authentication and safety research. The OECD describes it as "a common, voluntary structure for organisations to disclose their AI governance and risk management practices." Two words in that sentence do the work: common and voluntary.

What actually changed

The framework first went live in February 2025. The OECD calls it the first international, voluntary tool built to let organisations report against the Hiroshima Code of Conduct. An early round drew submissions from 20 organisations across different sectors and countries, and the OECD published a read of what those reports revealed about how AI developers say they manage risk.

In a March 2026 update, the OECD said it was streamlining and improving the framework and that a v2.0 "will be available shortly," with a pilot open at OECD.AI/Transparency. Secondary reporting later pointed to a v2.0 surfacing around a G7 ministerial in May 2026, but that timing rests on a press release we could not open, so treat it as unconfirmed. The direction of travel is simplification. A shorter template lowers the cost of filing, which is the OECD's lever for getting more organisations to participate. My read is that this is the whole point. A transparency register only becomes useful once enough names are in it to compare, and a heavy template keeps the list short.

Why a voluntary template still matters

Here is the practical tension. Filing is optional and there is no penalty for skipping it, so it is easy to file this under things you can ignore. I would not. A voluntary disclosure that a respected international body hosts tends to harden into a soft expectation. Once a handful of large developers publish, a customer or a government buyer can ask why you have not, and a competitor's report becomes a yardstick your own governance gets measured against. The framework says nothing about who must report, but the market reads absence.

There is also a drafting risk that senior people underrate. A HAIP report is a public statement about your controls. If what you write does not match what you actually do, you have created a document that a plaintiff's lawyer, a regulator, or an investigative reporter can hold up next to reality. The voluntary label protects you from a filing obligation. It does not protect you from your own words. Treat the report the way you would treat any public representation about your risk posture, because that is what it is.

What this means for US professionals

The Hiroshima process is a G7 initiative, and the reporting framework does not bind anyone in the United States. Its value to a US audience is as a shared vocabulary. If you advise or sit inside a company that builds or deploys advanced AI, the framework is a ready-made checklist of the governance topics that international policymakers now expect a serious operator to have answers on. You can use it to pressure-test your own program even if you never file. For counsel and compliance leaders, a published report from a peer is a free look at how comparable firms describe their controls, which is useful for benchmarking and for spotting where your own disclosures would look thin.

The judgment call is whether to file. That is a real decision with arguments on both sides, and it belongs with your legal and communications teams, not with a template. What a streamlined v2.0 would change is the cost of saying yes. It does not change the underlying calculus that a public governance statement is a commitment you have to be able to stand behind.

Questions professionals are asking

Is the Hiroshima reporting framework mandatory?

No. The OECD describes it as "a common, voluntary structure for organisations to disclose their AI governance and risk management practices." Nobody is required to file, and there is no penalty for organisations that choose not to. It is a transparency template tied to the G7's voluntary Code of Conduct, not a law.

What is different about the streamlined v2.0?

In a March 2026 update, the OECD said it was streamlining and improving the framework and that a simpler v2.0 "will be available shortly," with a pilot open at OECD.AI/Transparency. Some secondary reporting later pointed to a v2.0 around a G7 ministerial in May 2026, but that timing is unverified. The intended change is a shorter, simpler template to make it easier for more organisations to participate. The core purpose, voluntary disclosure against the Hiroshima Code of Conduct, is unchanged.

Does this apply to US companies?

Not as law. The Hiroshima AI Process is a G7 initiative, and the reporting framework binds no one in the United States. For US firms it works as a shared checklist of the governance topics international policymakers expect an advanced-AI developer to address, and as a benchmark against peers who publish reports.

Should we file a report?

That is a judgment call for your legal and communications teams. Filing is optional, but a published report becomes a public statement about your controls that others can hold against your actual practice, so it should be accurate and signed off before release. The streamlined v2.0 lowers the effort of filing; it does not change the fact that the report is a commitment.

RELATED BRIEFINGS

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any framework, standard, or code applies to your situation with qualified professionals in the relevant jurisdiction.