AI Safety Institute
New technical institute for AI evaluation and testing.
How the laws define this
Regulatory frameworks consistently position an AI Safety Institute as an institutional body tasked with evaluating systems, conducting testing, and providing expert evidence or guidance to support safe AI implementation. The Government response to the AI regulation white paper, the India AI Governance Guidelines: Enabling Safe and Trusted AI Innovation, and the Bletchley Declaration focus primarily on technical execution, requiring these entities to perform capability assessments, build test suites, and generate scientific metrics to support regulatory oversight. In contrast, the National AI Plan 2025 emphasizes an advisory role, tasking the institute with delivering expert advice to industry, agencies, and ministers on safe and ethical adoption rather than technical evaluation. Scope and operational details also diverge: the Bletchley Declaration explicitly includes cross-national entities assessing frontier AI systems through public–private partnerships, whereas the National AI Plan 2025 uniquely specifies a $29.9 million establishment budget and a 2026 commencement date.
Synthesised from the 4 statutory definitions below. Each one is quoted in full, with its source.
How jurisdictions differ
What is distinctive in each definition, binding law in force first. Showing 3 of the 4 instruments that define this term; all are listed under Definitions below.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| India | India AI Governance Guidelines: Enabling Safe and Trusted AI Innovation | In Force | Proposes an institute under IndiaAI Mission to develop test suites and benchmarks, conduct research on AI safety and resilience, and supply technical evidence directly to the AIGG for regulatory oversight. |
| United Kingdom | A pro-innovation approach to AI regulation: government response | Adopted | Establishes a UK body to build testing capacity, support regulator coordination with the central AI risk function, and inform risk assessments and technical standards. |
| United Kingdom | Bletchley Declaration (outcome of 2023 AI Safety Summit) | Adopted | Broadens to national or cross-national entities partnering with governments to provide testing infrastructure and support for state-led or joint public-private safety assessments of frontier AI. |
Definitions (4)
A newly established body tasked with providing technical evaluation, testing capacity, and evidence to support monitoring, assurance techniques, and regulator coordination. It will work closely with the central AI risk function and sectoral regulators to inform risk assessments and technical standards.
A proposed technical institute tasked with conducting safety testing, developing test suites and benchmarks, performing evaluation and research on AI safety and resilience, and providing technical evidence and metrics to support oversight by regulators and the AIGG.
Established with $29.9 million, commencing operations in 2026, mandated to provide expert advice to industry, agencies, and ministers on safe and ethical AI adoption.
National or cross-national entities partnering with governments to undertake capability assessments, safety testing, and evaluation of frontier AI systems, providing scientific expertise, testing infrastructure, and support for state-led or joint public–private safety assessments.
Not to be confused with
AI Security Institute focuses on AI security, capability testing, and red-teaming, while this term covers broader AI evaluation, testing, and safety assessments.
Australian AISI is an independent institute for AI safety analysis and testing, whereas this term describes entities providing evaluation support to regulators or governments.
Questions and answers
What is the role of the AI Safety Institute in India?
It conducts safety testing, develops test suites and benchmarks, performs evaluation and research on AI safety and resilience. It provides technical evidence and metrics to support oversight by regulators and the AIGG. This is part of the India AI Governance Guidelines under the IndiaAI Mission.
How does the UK define the AI Safety Institute?
It is a body providing technical evaluation, testing capacity, and evidence to support monitoring, assurance, and regulator coordination. It works with the central AI risk function and sectoral regulators to inform risk assessments and standards.
What does the Bletchley Declaration say about AI Safety Institutes?
It refers to national or cross-national entities partnering with governments for capability assessments, safety testing, and evaluation of frontier AI systems. These provide scientific expertise, testing infrastructure, and support for public-private safety assessments.
Related Terms
AI Safety Institute (AISI)
Technical body for safety testing, research and standards....
AI Safety Institutes
National or cross-national bodies for capability assessment and testing....
AI Security Institute (AISI)
UK expert body for AI security, capability testing and red‑teaming....
Australian Artificial Intelligence Safety Institute (AISI)
New independent institute for AI safety analysis and testing....
Government Center for AI Regulation and Policy
National coordination and knowledge centre for AI oversight....