The UK AI Safety Institute (UK AISI) was a state-backed organization established in November 2023 to evaluate and ensure the safety of advanced artificial intelligence (AI) models, particularly frontier AI systems. Its creation marked a significant step in the global movement toward formal AI safety regulation, following public declarations in 2023 about potential existential risks from AI. The institute evolved from the Frontier AI Taskforce and was announced by Prime Minister Rishi Sunak, who emphasized the need for independent safety evaluations, stating that AI companies cannot "mark their own homework." Initially based in London, the UK AISI later expanded to San Francisco in May 2024 to be closer to major AI developers. In 2025, it was renamed the AI Security Institute, reflecting a broader focus on security concerns.
Background and Motivation
The concept of an artificial intelligence safety institute emerged as a type of state-backed organization aimed at evaluating and ensuring the safety of advanced AI models. In 2023, AI safety gained prominence, with notable public statements from researchers and industry leaders about the potential existential risks posed by AI. The UK government, under Prime Minister Rishi Sunak, sought to position the country as a leader in AI safety regulation. Sunak expressed his intention to "make the UK not just the intellectual home but the geographical home of global AI safety regulation," and unveiled plans for an AI Safety Summit. The summit was held in November 2023 at Bletchley Park, a historic site known for its role in World War II codebreaking, symbolizing the UK's commitment to technological security.
Establishment and Evolution
The UK AISI was officially established during the AI Safety Summit in November 2023, evolving from the Frontier AI Taskforce, which had been created earlier that year. The institute's primary mission was to conduct safety evaluations of frontier AI models before their deployment, focusing on capabilities such as Large language model reasoning, Generative AI content generation, and potential misuse. The institute aimed to provide independent assessments, reducing reliance on AI companies' self-evaluations. In April 2024, the UK and the US concluded an agreement to collaborate on at least one joint safety test, marking a milestone in international cooperation. The UK AISI announced in May 2024 that it would open an office in San Francisco, where many leading AI companies, including OpenAI, Anthropic, and Google DeepMind, are headquartered. This move was part of a plan to "set new, international standards on AI safety," according to UK technology minister Michele Donelan.
International Network
The creation of the UK AISI catalyzed the formation of a global network of AI safety institutes. At the AI Seoul Summit in May 2024, international leaders agreed to form a network comprising institutes from the UK, the US, Japan, France, Germany, Italy, Singapore, South Korea, Australia, Canada, and the European Union. This network aimed to harmonize evaluation procedures and share best practices. In July 2025, the network held an exercise to explore issues with evaluating AI agents, particularly concerning leaking sensitive information or cybersecurity. Members also met at NeurIPS 2025 in San Diego, a major conference for Machine learning research, to discuss progress and future directions.
Specific Institutes
Australia
The Albanese government announced the creation of the Australian AI Safety Institute on 25 November 2025. The institute is housed within the Department of Industry, Science and Resources and is supported by a budget of A$29,900,000 over four years. Its general manager is Kate Conroy, who also leads responsible AI initiatives in the Royal Australian Air Force.
Canada
Canada announced in April 2024 that it would create an AI safety institute, which was officially founded in November 2024. The institute operates under Innovation, Science and Economic Development Canada, partnering with the Canadian Institute for Advanced Research (CIFAR). It is supported by a budget of CA$50,000,000 over five years.
European Union
The EU AI Office, founded in May 2024, is a member of the international network of AI safety institutes. It coordinates AI safety efforts across member states, focusing on regulation and standards.
France
On 31 January 2025, the French government created the Institut national pour l'évaluation et la sécurité de l'intelligence artificielle (INESIA), or the National Institute for AI Evaluation and Security. This institute aims to evaluate AI systems and contribute to European and international safety standards.
India
The Ministry of Electronics and Information Technology held consultations with various stakeholders, including meta-platforms (Meta), Google, Microsoft, IBM, OpenAI, NASSCOM, and Indian Institutes of Technology (IITs), on October 7, 2024, regarding the establishment of an AI Safety Institute. The decision was made to shift focus from regulation to standards-setting, risk identification, and damage detection. Minister Ashwini Vaishnaw announced the creation of an IndiaAI Safety Institute on January 30, 2025, to ensure ethical and safe AI applications. The institute will promote domestic R&D grounded in India's social, economic, cultural, and linguistic diversity, using Indian datasets. It operates under a "hub-and-spoke" model with academic and private sector partners, including IITs and international organizations like UNESCO.
Japan
The Japan AISI (J-AISI) was founded in February 2024 as part of the Information Technology Promotion Agency. It employs about 23 people and consists of the Council of AISI, the AISI Steering Committee, and a secretariat with six teams. Akiko Murakami, previously of IBM Japan and Sompo Japan, serves as executive director, with Kenji Hiramoto and Suguru Nishimura as deputy executive directors.
Kenya
Kenya agreed to join the international network of AI safety institutes, becoming the only African state in the network, though no specific details have been announced yet.
Singapore
The Digital Trust Centre, initially founded in June 2022, was renamed the Singapore AISI in May 2024. It is part of Nanyang Technological University and partners with the Infocomm Media Development Authority, supported by an investment of S$10,000,000 per year.
South Korea
South Korea announced in May 2024 that it would create an AI safety institute under the Electronics and Telecommunications Research Institute. The institute was founded in November 2024, based in Bundang District, with a tentative investment of 10 to 20 million South Korean won per year and at least 30 employees.
Impact and Legacy
The UK AI Safety Institute's founding set a precedent for state-backed AI safety evaluation, influencing the creation of similar bodies worldwide. Its evolution into the AI Security Institute in 2025 reflected a broadening scope to include security concerns beyond safety. The institute's work has contributed to the development of international standards, though challenges remain, as noted by Politico in April 2024, when many AI companies had not shared pre-deployment access to their most advanced models for evaluation. Meta's president of global affairs, Nick Clegg, noted that companies were waiting for common evaluation rules and procedures, which the UK and US institutes worked to establish. The international network continues to play a crucial role in coordinating AI safety efforts, with regular exercises and meetings to address emerging challenges in Artificial intelligence governance.