AI Security Institute
 | |
| Formation | November 2023 |
|---|---|
| Services | AI safety research |
Parent organization | Department of Science, Innovation and Technology |
| Website | www |
The AI Security Institute (AISI) is a research organisation under the Department for Science, Innovation and Technology of the United Kingdom that aims "to equip governments with a scientific understanding of the risks posed by advanced AI".[1] It conducts research, and develops and tests mitigations.
The organisation was created as the AI Safety Institute and renamed in February 2025.[2] Its creation followed the world's first major AI Safety Summit that was held in Bletchley Park in 2023.[3] The institute's stated goal is "building the world's leading understanding of advanced AI risks and solutions, to inform governments so they can keep the public safe". It is designed like a startup in the government "combining the authority of government with the expertise and agility of the private sector".[4]
AISI has made access agreements with Anthropic, Google and OpenAI to test their models before release.[3] It has an open source platform called Inspect that permits companies, governments and academics to run standardised safety tests for AI usage.[3] Among its work is the reported detection of multiple serious vulnerabilities that could enable development of biological weapons; the vulnerabilities were fixed before the model was launched.[3]
It conducts research on diverse fields of AI application. One study by AISI found that LLMs post-trained for political persuasiveness became systematically less accurate and up to 51% more persuasive on political issues.[5] AISI has also worked on the usage of AI for emotional needs. It found that nearly 10 percent of UK citizens used systems like chatbots for emotional purposes on a weekly basis.[6] It found that "systems are now outperforming PhD-level researchers on scientific knowledge tests and helping non-experts succeed at lab work that would previously have been out of reach" in a report published in December 2025.[7]
Role in AI Governance
[edit]The AI Security Institute is part of the wider AI governance area in that it offers scientific research and evaluation of advanced AI technologies that will ensure evidence-based policies. With the institute focusing on the scientific methods of evaluating frontier AI technologies, it ensures the creation of sound evidence that will help in decision-making by the government and other stakeholders regarding AI technology[8].
A critical part of the institute’s governance role includes the creation and utilisation of methods of rigorous assessment of advanced AI systems. The standardization of test procedures, safety standards, and model assessments enhances the objectivity and comparability of the assessment of AI, helping to contribute to the general process of establishing best practices to govern the potential risks of advanced AI technologies [9]. By contributing technical facts, not regulatory policies, the institute provides policymakers and public officials with information that will be useful in developing governance models for advanced AI [8].
In addition to this, the institute is contributing to AI governance through collaborations between governments, researchers from academia, and AI developers. Collaboration in this way helps to share knowledge in technology, adopt common practices in safety evaluation, and coordinate on risk response measures that transcend national borders. Collaborative efforts of this nature have been found to be an essential component of good AI governance for advanced AI technologies since several safety and security issues demand common scientific understanding and interoperable evaluation practices [8].
Such tasks correspond with the general international guidelines regarding responsible governance of AI, which focus on principles of transparency, accountability, risk management, and reliance on scientific evidence in decision-making [10]. In the same way, the frameworks for AI governance encourage businesses to apply a structured approach to the identification, assessment, and management of the associated risks through the entire life cycle of AI [9]. In addition, recent international policy recommendations emphasize the significance of independent evaluation, cross-sector cooperation, and strong technical expertise in facilitating the responsible development and deployment of frontier AI technologies[11] . Altogether, the set of tasks makes the AI Security Institute a valuable participant in the emerging international system of AI governance alongside its research agenda, facilitating the gap between technical AI safety research and public-sector governance[12] .
Former chief AI officer of GCHQ Adam Beaumont is the institution's interim director. Jade Leung is the chief technology officer,[4] who in August 2025 was appointed as the prime minister's AI adviser.[13]
See also
[edit]- Alan Turing Institute – UK institute for data science and AI
- Artificial intelligence safety institute – worldwide topic
References
[edit]- ↑ "AISI". Government of UK. Retrieved 20 February 2026.
- ↑ "Written statements - Written questions, answers and statements - UK Parliament". questions-statements.parliament.uk. Retrieved 2026-03-23.
- 1 2 3 4 "Britain is the closest the world has to an AI safety inspector". The Economist. ISSN 0013-0613. Retrieved 2026-02-20.
- 1 2 "Equipping governments with a scientific understanding of the risks posed by advanced AI". AI Security Institute.
- ↑ "The levers of political persuasion with conversational artificial intelligence". Science. 2025-12-04.
- ↑ Milmo, Dan (18 December 2025). "Third of UK citizens have used AI for emotional support, research reveals". The Guardian. ISSN 0261-3077. Retrieved 2026-02-20.
- ↑ AISI. Frontier AI Trends Report. December 2025. retrieved on 20 February 2026
- 1 2 3 "International AI Safety Report". internationalaisafetyreport.org. 2026-02-03. Retrieved 2026-08-10.
- 1 2 "AI Risk Management Framework". NIST. 2021-07-12.
- ↑ "AI Principles Overview". oecd.ai. Retrieved 2026-08-10.
- ↑ "WEF AI Governance Summary". regulations.ai. Retrieved 2026-08-10.
- ↑ unesdoc.unesco.org https://unesdoc.unesco.org/ark:/48223/pf0000381137. Retrieved 2026-08-10.
{{cite web}}: Missing or empty|title=(help) - ↑ "Appointment of Jade Leung as the Prime Minister's AI adviser". GOV.UK. DSIT. 15 August 2025. Retrieved 6 August 2026.
