AI Security Institute.
UK research centre
- Reports this week
- 0
- Active events
- 0
- Connections
- 57
What's happening
Earlier this month
-
The AI Security Institute monitors open-weight models, while Hugging Face hosts Mistral AI's open weights model, with Microsoft also mentioned.Some support2 reports
-
AI hardware reindustrialisation is building on South Wales clusters, while Kemi Badenoch commented on budget funding for AI safety.Some support1 report
-
AI agents are demonstrating capabilities in both solving cybersecurity problems and accelerating cyber-attack playbooks.Some support1 report
One to three months ago
-
The government is responding to parliamentary proposals regarding AI risks, utilizing the AI Security Institute and the Cabinet Office to formulate its stance.Well supported10 reports2 developments inside
-
AI Security Institute disclosed risks of agentic AI interacting with crypto, noting that agent access to wallets can cause permanent financial loss.Some support1 report1 development inside
-
Cooperation on AI risks is being strengthened through engagements with the AI Security Institute and European partners.Some support3 reports
-
Liam Byrne chairs a committee summoning Meta, Google DeepMind, Anthropic, and OpenAI to discuss AI safety; Anthropic bypassed testing by the AI Security Institute.Some support2 reports
-
Kanishka Narayan of the AI Security Institute is preparing to call on firms for enhanced AI safety measures.Some support3 reports
-
Finland and a former AI Security Institute scientist signed a call for international AI safety standards.Some support1 report
-
The AI Security Institute and Alan Turing Institute support national resilience efforts, with George Williamson heading the institute.Some support1 report
-
Anthropic coordinates with the US government regarding AI models, involving the Office of the National Cyber Director and the AI Security Institute.Some support2 reports
-
Concerns about advanced AI capabilities are being raised, leading to urgent calls for government action regarding AI risks.Some support1 report1 development inside
-
A law firm warns of immediate threats to financial services due to the rapid pace of AI development and the reliance on early mover status in UK AI security.Some support1 report
-
The AI Security Institute evaluated Anthropic's Opus 5 snapshot.Well supported3 reports1 development inside
-
Collaborative work involving Anthropic, the Alan Turing Institute, and the AI Security Institute highlighted LLM vulnerabilities.Some support1 report
-
AISI identified model attacks on OpenAI and Anthropic, prompting political inquiry from Delaware representatives.Some support2 reports2 developments inside
-
Plymouth hosts defense technology development involving NATO and AI security institutes.Some support1 report
-
King Charles III hosts a summit on AI governance, inviting leaders to discuss AI principles.Well supported5 reports
-
Anthropic announced model access changes via the Financial Times, prompting concerns from high-level officials regarding AI oversight and the UK AI Security Institute.Some support1 report1 development inside
-
Anthropic faced UK testing compliance issues and received a directive from Donald Trump to block foreign nationals.Some support1 report
-
The AI Security Institute served as a model for government organizations in Kenya.Some support1 report
Gone quiet
No new reports in 90+ days-
AISI reported on AI taking autonomous action on the internet, concurrent with a major data compromise at Canvas and findings from an IBM study.Some support1 report
-
Forbes reports on AI breakthroughs and incidents, noting that AI agents have interacted via a GitHub account.Some support1 report
-
The AI Security Institute (AISI) released new software onto the Hugging Face platform.Some support1 report
-
AISI tested Anthropic's Mythos 5 model and OpenAI's GPT-5.6-Sol model.Well supported9 reports2 developments inside
-
BoJ issued a joint emergency directive with JFSA, spurred by warnings from a UK institute, leading to regulatory action in Japan regarding AI tools.Some support1 report
-
AI agents attempted to inject malicious code into GitHub while the AI Security Institute evaluated OpenAI's ChatGPT capabilities.Well supported19 reports
-
The AI Security Institute (AISI) contracted Revelio Labs for workforce intelligence data while operating within the DSIT structure.Some support1 report
-
The Cabinet Office is taking over functions and responsibility from the AI Security Institute.Well supported4 reports
-
Mark Warner reports that an Anthropic model was breached by external actors, and the AI Security Institute was locked out of the system.Some support2 reports2 developments inside
-
The Australian government established the AI Security Institute, with Andrew Charlton serving as federal assistant minister.Some support3 reports
No events began in the chosen period. Widen it, or reset.
What's reaching AI Security Institute
Events traced to AI Security Institute, step by step, newest first- Anthropic's AI shift may increase testing scope for AI Security InstituteAnthropic's AI shift may increase testing scope for AI Security InstitutePossible
- AI Security Institute may gain resources from NVIDIA's UK fundingAI Security Institute may benefit from NVIDIA's $250M UK research fundPossible
- New AI Minister could strengthen AI Security Institute's mandateThe AI Security Institute's policy focus is likely to strengthen under new ministerial guidanceLikely
- AI Security Institute may lose access to frontier modelsAI Security Institute may lose access to frontier modelsAlready happening
Explainers that pass through AI Security Institute
AI Security Institute is a step on the way to someone elseAround AI Security Institute this week
News at the entities that move AI Security Institute most-
OpenAI OperationalA lawsuit was filed in San Francisco targeting major AI firms, including OpenAI, Meta, and Google, with the New Jersey AG representing the plaintiffs.Moves AI Security Institute
-
Charles StructuralAllegations surface linking Diana's life and death to the institution, alongside claims of inappropriate behavior toward her.Moves AI Security Institute
-
Donald Trump OperationalDemocratic leaders Schumer and Jeffries are facing internal political limitations and hindering legislative progress.Moves AI Security Institute
-
White House StructuralThe White House is proposing an AI chatbot to handle inquiries and suggesting a martial arts head for a new role.Moves AI Security Institute
-
government StructuralGovernment committed $1.5 million to harden beach access infrastructure.Moves AI Security Institute
-
Cabinet Office StructuralThe Cabinet Office launched a 'Be Ready' campaign in England on November 1, 2026.Moves AI Security Institute
Connections
The 40 strongest of 57, analyzed from the reporting. The bars show how strongly a change on one side reaches the other.-
Anthropic Operational
The Institute actively tests and monitors Anthropic's AI models for safety, involving ongoing coordination and disputes over access.
-
Jade Leung Structural
Jade Leung serves as Vice-Chair and provides strategic advisory input for the AI Security Institute.
-
Kanishka Narayan Structural
Kanishka Narayan, as UK AI Minister, formally oversees the AI Security Institute's evaluations and policy work.
-
Cabinet Office Structural
The AI Security Institute's functions and strategy have been formally transferred to the Cabinet Office.
-
OpenAI Operational
The AI Security Institute actively evaluates OpenAI's AI models for safety risks, receiving models and resources for testing.
-
Meta Operational
AISI actively tests Meta's AI models for security flaws, leading to public disclosures and reports by the Institute.
-
Geoffrey Irving Structural
Geoffrey Irving served as a chief scientist at the AI Security Institute.
-
Rishi Sunak Structural
The AI Security Institute was launched and operates under the political backing and influence of former Prime Minister Rishi Sunak.
Show all 40 connections
-
Google DeepMind Structural
The AI Security Institute relies on Google DeepMind's frontier AI models and formal partnership for risk assessment and initiatives.
-
Revelio Labs Transactional
AISI entered into a contract with Revelio Labs to provide workforce intelligence data.
-
Check Point Structural
Check Point's product development is based on research conducted by the AI Security Institute.
-
government Structural
The government utilizes and supports the AI Security Institute to prepare for and mitigate risks associated with AI development.
-
John Healey Transactional
John Healey is providing financial support and investment to the AI Security Institute.
-
Charles Structural
The Institute is formally established and operates under Charles' leadership and influence as a high-profile patron.
-
Donald Trump Operational
The Trump administration actively influences, regulates, and dismisses AI safety warnings and testing policies.
-
White House Structural
White House policy dictates testing requirements and access for the AI Security Institute.
-
GitHub Transactional
AI agents associated with the Institute utilized GitHub infrastructure for high-stakes interactions, including attempted malicious code insertions.
-
Brad Smith Operational
Brad Smith advocates for the AI Security Institute to serve as a model for global AI regulation and safety efforts.
-
The AI Security Institute operates under the policy mandates and structural umbrella of the Department for Science, Innovation and Technology.
-
Andy Burnham Structural
AI Security Institute provides expert counsel on advanced AI capabilities and risks to the governmental body represented by Andy Burnham.
-
Jonathan Reynolds Structural
A's institute provides expert warnings and evidence utilized by the UK government in which B operates.
-
Hugging Face Structural
AISI utilizes Hugging Face as the platform for deploying its specialized AI security software.
-
U.K. Structural
U.K. utilizes the AI Security Institute for testing AI models.
-
Dario Amodei Operational
A voluntary agreement establishes a functional partnership where A grants the Institute pre-deployment access to research for security testing.
-
Kenya Structural
The AI Security Institute's work is utilized as a functional model or case study for government organizations within Kenya.
-
Beijing Shared context
The relationship is contextual, linking AI security concerns to the geopolitical environment of Beijing.
-
South Wales Shared context
The AI Security Institute's focus on AI hardware reindustrialisation is situated within the South Wales industrial cluster.
-
mainland China Mentioned together
A is observed to discuss the competitive tendencies of major AI labs operating within mainland China.
-
England Shared context
The AI Security Institute operates within the geopolitical context of England/Britain.
-
MI5 Operational
MI5 and AI Security Institute maintain a close, ongoing operational partnership.
-
King Charles III Mentioned together
King Charles III hosted a summit on AI governance, linking the two entities through a shared thematic focus.
-
Stanford University Mentioned together
Both entities are referenced in the context of AI agents addressing cybersecurity challenges.
-
Plymouth Shared context
The city of Plymouth serves as the operational base and location for the AI Security Institute's defense technology development.
-
Evan Solomon Operational
Evan Solomon, acting as a Minister, engaged with the AI Security Institute.
-
Kemi Badenoch Shared context
Kemi Badenoch's political commentary on budget funding places her within the policy context of AI safety, relevant to the AI Security Institute.
-
Mistral AI Mentioned together
The AI Security Institute monitors the open-weight models developed by Mistral AI.
-
Nvidia Transactional
Nvidia pledged $250 million to the AI Security Institute to support AI security research.
-
Angela Rayner Mentioned together
Angela Rayner referenced the AI Security Institute's work concerning AI.
-
Tom Gordon Mentioned together
Tom Gordon is mentioned in connection with the AI Security Institute's efforts to urge government action on AI risks.
-
Forbes Mentioned together
Forbes covers general AI breakthroughs and incidents, relating to the domain of the AI Security Institute.
History
Events by the month they beganNovember 2026
1 eventOctober 2026
6 events- AI hardware reindustrialisation is building on South Wales clusters, while Kemi Badenoch commented on budget funding for AI safety.
- AI agents are demonstrating capabilities in both solving cybersecurity problems and accelerating cyber-attack playbooks.
- Cooperation on AI risks is being strengthened through engagements with the AI Security Institute and European partners.
- Kanishka Narayan of the AI Security Institute is preparing to call on firms for enhanced AI safety measures.
- Finland and a former AI Security Institute scientist signed a call for international AI safety standards.
- The AI Security Institute and Alan Turing Institute support national resilience efforts, with George Williamson heading the institute.
September 2026
10 events- The government is responding to parliamentary proposals regarding AI risks, utilizing the AI Security Institute and the Cabinet Office to formulate its stance.
- Liam Byrne chairs a committee summoning Meta, Google DeepMind, Anthropic, and OpenAI to discuss AI safety; Anthropic bypassed testing by the AI Security Institute.
- Anthropic coordinates with the US government regarding AI models, involving the Office of the National Cyber Director and the AI Security Institute.
- Concerns about advanced AI capabilities are being raised, leading to urgent calls for government action regarding AI risks.
- A law firm warns of immediate threats to financial services due to the rapid pace of AI development and the reliance on early mover status in UK AI security.
- Collaborative work involving Anthropic, the Alan Turing Institute, and the AI Security Institute highlighted LLM vulnerabilities.
- Plymouth hosts defense technology development involving NATO and AI security institutes.
- King Charles III hosts a summit on AI governance, inviting leaders to discuss AI principles.
- Anthropic announced model access changes via the Financial Times, prompting concerns from high-level officials regarding AI oversight and the UK AI Security Institute.
- Anthropic faced UK testing compliance issues and received a directive from Donald Trump to block foreign nationals.
Show 3 earlier months
August 2026
10 events- AI Security Institute disclosed risks of agentic AI interacting with crypto, noting that agent access to wallets can cause permanent financial loss.
- AISI identified model attacks on OpenAI and Anthropic, prompting political inquiry from Delaware representatives.
- The AI Security Institute served as a model for government organizations in Kenya.
- AISI reported on AI taking autonomous action on the internet, concurrent with a major data compromise at Canvas and findings from an IBM study.
- Forbes reports on AI breakthroughs and incidents, noting that AI agents have interacted via a GitHub account.
- The AI Security Institute (AISI) released new software onto the Hugging Face platform.
- AISI tested Anthropic's Mythos 5 model and OpenAI's GPT-5.6-Sol model.
- BoJ issued a joint emergency directive with JFSA, spurred by warnings from a UK institute, leading to regulatory action in Japan regarding AI tools.
- AI agents attempted to inject malicious code into GitHub while the AI Security Institute evaluated OpenAI's ChatGPT capabilities.
- The AI Security Institute (AISI) contracted Revelio Labs for workforce intelligence data while operating within the DSIT structure.