12 Companies Published Frontier AI Safety Frameworks in 2025, International AI Safety Report Finds
What happened
The International AI Safety Report published its 2026 Report: Extended Summary for Policymakers on May 9, 2026, providing a policymaker-focused synthesis of the global frontier AI safety landscape. A central finding of the report is that 12 companies published or updated Frontier AI Safety Frameworks in 2025, documents that outline how those organizations identify, evaluate, and manage risks associated with the development of advanced AI systems. The report takes a global scope rather than focusing on a single jurisdiction, positioning it as a reference point for regulators and compliance functions across multiple legal systems simultaneously. It contextualizes these frameworks within parallel international efforts including the Bletchley Declaration on AI Safety and the Singapore Consensus on Global AI Safety Research Priorities. As an authoritative account from a recognized international body, the report is expected to inform regulatory proposals, parliamentary inquiries, and multilateral negotiations in the months following its release.
Why it matters
- ·Regulatory exposure is heightened because the aggregation of 12 Frontier AI Safety Frameworks in a single authoritative document gives bodies such as the EU AI Office, the UK AI Safety Institute, and national legislatures a concrete benchmark for identifying gaps and inconsistencies, accelerating the transition from voluntary commitments toward mandatory disclosure requirements such as those emerging under the EU AI Act.
- ·Operational impact is immediate for organizations developing or deploying advanced AI systems, as legal and risk teams must now benchmark their internal AI risk management documentation against the 12 published frameworks before mandatory disclosure regimes come into effect in key jurisdictions.
- ·Organizational risk extends to procurement and vendor management functions, because the report equips policymakers with a basis for distinguishing AI developers with documented safety commitments from those without, which is likely to influence procurement standards and contractual due diligence requirements across regulated industries.
Governance controls affected
What to do now
- ☐Conduct a gap analysis comparing your organization's existing internal AI risk management documentation against the 12 Frontier AI Safety Frameworks documented in the report, and log findings in your model card and documentation maintenance records.
- ☐Assess whether your organization meets the threshold for publishing a Frontier AI Safety Framework, treating the 12 documented examples as a reference set that regulators and counterparties are now consulting when evaluating risk posture.
- ☐Review your AI risk classification inventory to ensure that frontier and advanced AI systems are correctly categorized and that associated risk management plans are audit-ready ahead of potential mandatory disclosure requirements under the EU AI Act or the California Transparency in Frontier AI Act.
- ☐Update third-party AI risk assessment procedures and vendor contract requirements to reflect the report's findings, ensuring that AI developers in your supply chain can demonstrate documented safety commitments aligned with recognized international frameworks.
- ☐Schedule a tabletop exercise or internal review to test whether your incident disclosure and notification procedures satisfy the transparency expectations set by the Bletchley Declaration and referenced in the report.
What to watch next
Compliance teams should monitor the EU AI Office and UK AI Safety Institute for guidance or enforcement signals that cite the 2026 Report findings as a baseline standard for frontier AI risk management obligations. Progress on mandatory framework requirements under the EU AI Act and the California Transparency in Frontier AI Act should be tracked closely, as regulators may reference the 12 documented frameworks when setting minimum disclosure thresholds. Teams should also watch for follow-on outputs from the Singapore Consensus process and any updates to the Bletchley Declaration commitments, as these multilateral instruments are likely to generate new technical benchmarks and compliance expectations for organizations developing advanced AI systems.