On October 14, 2026, Demis Hassabis called for an international body to set standards for safe AI development and use. At the Dublin International AI Summit, the Google DeepMind co-founder and president also advocated a flexible certification process that would adjust to detected risks and changes in the technology, according to EFE’s report published by Diario de León. NeoTeo previously covered another Google DeepMind initiative, the DeepMind Institute.
The idea is a proposed international governance body and a risk-proportional approach to certification. Two existing initiatives serve different purposes: ISO/IEC 42001:2023 sets requirements for an organization’s AI management system, while the International AI Safety Report 2026 synthesizes scientific evidence to inform policymaking.
What Hassabis proposed
Hassabis called for an international body to set standards for the safe development and use of AI. He described certification that would be dynamic, proportionate to identified risks, and able to adapt as the technology evolves.
He linked the call to cybersecurity concerns involving AI agents. In EFE’s account, the examples included attempts to gain unauthorized access to organizations’ systems, website manipulation to obtain restricted information, and a false police lead during an automated test. Hassabis characterized recent incidents as early warning signs and warned that more serious incidents could occur in the next year or two if the AI sector, individual companies, and governments did not strengthen safety efforts.
What existing international initiatives do
ISO/IEC 42001:2023 is a published international standard for establishing, implementing, maintaining, and continually improving an organization’s AI management system. It focuses on organizational policies and processes, rather than technical requirements for each AI application. Organizations may choose independent certification against the standard to confirm that their management system meets its requirements. The standard was published on December 18, 2023, by the International Organization for Standardization.
The International AI Safety Report 2026 has a different role: it brings together scientific evidence about general-purpose AI capabilities, risks, and risk management to inform policymakers. It does not make specific policy recommendations. The second edition drew contributions from more than 100 experts, with an Expert Advisory Panel nominated by more than 30 countries and international organizations, according to the report’s official page.
| Initiative | Purpose and subject | Instrument type |
| Hassabis’s proposal | International standards for safe AI development and use, with adaptable certification proportionate to detected risks | Proposed international body and certification approach |
| ISO/IEC 42001:2023 | An organization’s AI management system | Published international management-system standard; independent certification is optional |
| International AI Safety Report 2026 | Scientific evidence on AI capabilities, risks, and risk management for policymakers | Expert assessment, not a policy or certification standard |
