The UK government announced its first external advisory board for the Frontier AI Taskforce on September 7, 2023. The announcement also renamed the former AI Foundation Model Taskforce, added senior AI researchers and security officials to its advisory structure, and outlined plans to build independent government capacity to assess the risks of highly capable AI systems.
The seven initial advisers were Yoshua Bengio, Paul Christiano, Matt Collins, Anne Keast-Butler, Alex van Someren, Helen Stokes-Lampard and Matt Clifford, who became vice-chair. The announcement was part of preparations for the UK AI Safety Summit held at Bletchley Park on November 1–2, 2023.
As an Amazon Associate I earn from qualifying purchases.
What the UK government announced
The September 7 announcement was more than a personnel change. The government combined four developments:
- The AI Foundation Model Taskforce was renamed the Frontier AI Taskforce.
- The taskforce appointed its first seven-member external advisory board.
- Matt Clifford was named vice-chair.
- The taskforce announced further technical appointments, research collaborations and preparations for the UK’s forthcoming AI Safety Summit.
The taskforce was created earlier in 2023 with £100 million in government backing. Its focus was not general AI regulation or routine automation. It was intended to help the government understand and evaluate risks from the most capable foundation models—systems whose capabilities could potentially create serious public-safety, national-security or global-security risks.
#1 Best Overall
The government’s first progress report described the taskforce as a rapidly built “startup inside government.” Its stated aim was to develop technical expertise and the ability to test advanced models independently, rather than relying entirely on AI companies to assess their own systems.
Who joined the advisory board?
The members brought together technical AI researchers, national-security officials, cybersecurity leadership, scientific advisers and a healthcare representative. Their roles below reflect their positions and expertise at the time of the September 7, 2023 announcement.
| Member | Role or expertise at the time | Why the appointment mattered |
|---|---|---|
| Yoshua Bengio | Deep-learning pioneer, 2018 A.M. Turing Award recipient, professor at Université de Montréal, and founder and scientific director of Mila. | Added internationally recognised expertise in AI research and safety. |
| Paul Christiano | Co-founder of the Alignment Research Center and former head of language-model alignment at OpenAI. | Added technical experience in AI alignment, control and the problem of making advanced systems behave as intended. |
| Matt Collins | UK Deputy National Security Adviser for Intelligence, Defence and Security. | Connected frontier-AI work to national-security policy. |
| Anne Keast-Butler | Director of GCHQ. | Added senior intelligence, cyber-risk and national-security expertise. |
| Alex van Someren | UK Chief Scientific Adviser for National Security and a former deep-tech investor and entrepreneur. | Added scientific assessment and technology-development experience. |
| Helen Stokes-Lampard | Physician and medical-sector leader; chair of the Academy of Medical Royal Colleges at the time. | Provided a healthcare and frontline public-service perspective on how advanced AI could affect society. |
| Matt Clifford | Prime Minister’s joint representative for the AI Safety Summit, chair of ARIA and co-founder of Entrepreneur First. | Served as vice-chair and linked the taskforce to the upcoming international summit. |
The composition illustrated the government’s view that frontier AI was simultaneously a technical, economic, social and security issue. It was not simply a panel of academic computer scientists, nor a conventional intelligence committee.
Why combine AI researchers with security officials?
Advanced AI models can create risks in areas that do not fit neatly into one government department. The taskforce identified concerns involving:
- cybersecurity and the discovery or exploitation of vulnerabilities;
- biosecurity and the potential misuse of biological information;
- public safety and dangerous model capabilities;
- national security and the effect of highly capable systems on defence and intelligence;
- social and institutional harms; and
- the reliability, controllability and alignment of advanced models.
Technical researchers can investigate what a model can do and whether it can be controlled. Security officials can assess how those capabilities might interact with hostile actors, critical infrastructure or state security. Healthcare and public-service expertise can help identify consequences that may not be visible from a purely technical or defence perspective.
Rank #2
This was also the rationale for developing public-sector evaluation capacity. The government wanted a neutral body capable of testing frontier models for dangerous or unexpected capabilities. At the time, that capability was still being built; the announcement should not be read as proof that a fully mature independent evaluation regime already existed.
What does “frontier AI” mean here?
In this policy context, “frontier AI” referred to highly capable foundation models, rather than every AI product or machine-learning system. The concern was that future or existing models could acquire capabilities sufficiently dangerous to threaten public safety or global security.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteThat distinction matters. A narrow fraud-detection model, an ordinary business-automation tool or a standard generative-AI application may raise important regulatory questions, but they are not automatically “frontier AI” under the taskforce’s definition. The taskforce was concentrating on the upper end of capability and the possibility of novel, difficult-to-predict risks.
What authority did the board have?
The advisory board was not an independent regulator, a statutory tribunal or a body with binding decision-making powers. Its role was to provide evidence-based advice to the taskforce and contribute expertise to its work.
The published terms said that members served as individuals, not as representatives of their employers. They were expected to participate actively, disclose conflicts of interest and withdraw from procurement discussions involving actual or perceived conflicts. Members were also not meant to use the board to promote company-specific work.
Those safeguards were important because the board included people connected to government, major research institutions, technology companies and AI-safety organisations. Formal conflict rules, however, should not be confused with independent evidence about how much influence the board ultimately had or how its recommendations were applied.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11The taskforce was also building an internal research team
The board appointments were separate from the taskforce’s employee and research appointments. The progress report said:
- Yarin Gal would join from Oxford as research director.
- David Krueger of the University of Cambridge would advise in a consultative role while the research programme was scoped.
- Ollie Ilott became taskforce director.
The taskforce said that, after roughly 11 weeks of recruitment, its researchers had accumulated more than 50 years of collective frontier-AI research experience. That figure was a claim in the taskforce’s own progress report, not an independently audited measurement. The report also said staff had experience at organisations including DeepMind, Microsoft, Redwood Research, the Center for AI Safety and the Center for Human-Compatible AI.
The government’s intended model was therefore broader than an advisory committee. It was trying to create an in-house technical capability, gain access to advanced models and compute, and work with external specialists on evaluation and risk research.
Which organisations worked with the taskforce?
The first progress report listed partnerships or collaborations involving:
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →- ARC Evals;
- Trail of Bits;
- The Collective Intelligence Project;
- the Center for AI Safety;
- Redwood Research; and
- RAND.
Later government material also described work involving Advai, Gryphon Scientific and Faculty AI. These relationships did not necessarily have identical contractual status or scope. Some were described as partnerships or collaborations, while later work included specific contracts. The distinction matters when assessing how much capability the taskforce controlled directly.
A later government announcement provides additional detail on organisations brought in to research frontier-AI risks.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How the appointments fitted the UK’s AI strategy
The appointments came during an unusually concentrated period of UK AI policy development. The government had announced the £100 million taskforce earlier in 2023 and said the UK would host the first international AI Safety Summit at Bletchley Park on November 1–2, 2023.
Matt Clifford’s role connected the taskforce directly to that summit. The board was intended to help the government prepare for discussions about how countries should evaluate advanced AI, share information about serious risks and coordinate internationally.
Recommended Free Tools
The announcement did not create the UK’s first AI advisory structure. The government had previously used bodies including the AI Council and the Centre for Data Ethics and Innovation. The Frontier AI Taskforce represented a more specific push toward technical risk assessment and frontier-model safety.
Best Value
Questions about representation and accountability
The board’s composition offered clear advantages: it combined leading AI expertise with officials who understood cyber, intelligence and national-security risks. It also included healthcare leadership, recognising that advanced AI could affect public services and professional practice.
But technical and security expertise does not automatically provide broad public representation. The board was not a citizens’ panel, labour representative body or general-purpose democratic forum. Reasonable questions included:
- Were civil-society, labour and consumer perspectives sufficiently represented?
- How much practical influence did the advisory board have over policy decisions?
- Could national-security priorities outweigh social, economic or democratic concerns?
- Were conflicts of interest and procurement decisions transparent enough?
- Were advice, meeting records and recommendations published?
These are governance questions, not established findings of misconduct or failure. The government’s published terms supplied formal safeguards, but they did not by themselves demonstrate how those safeguards operated in practice.
Free tools Windows power users keep installed
One-click scans. No signup required.
There was also a trade-off between speed and transparency. Rapid recruitment could help the UK build expertise before the Bletchley summit, but a fast-moving “startup inside government” can make selection criteria, accountability arrangements and public reporting harder to assess.
What happened next?
The Frontier AI Taskforce was a time-limited institutional stage rather than the UK’s final AI-safety structure. After the November 2023 AI Safety Summit, it evolved into the AI Safety Institute. Other AI-policy functions remained within the Department for Science, Innovation and Technology.
The subsequent change explains why readers looking for the Frontier AI Taskforce today may find references to the AI Safety Institute instead. It does not change what was announced on September 7, 2023: the government renamed the taskforce and created its first external advisory board to connect frontier-AI research with national-security and wider public-interest perspectives.
For the wider institutional timeline, the UK Parliament’s Communications and Digital Committee report records the taskforce’s renaming, the Bletchley summit and its later transition.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
