Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, whether US officials raised security concerns regarding Deepseek with her prior to the US announcement dated 19/10/2026.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
The United Kingdom and the United States share a strong partnership rooted in our shared values and people-to-people links, and demonstrated by our close collaboration on security, defence, trade and technology.
Ministers and officials have regular discussions with US counterparts on a range of technology and security issues. However, details of confidential government-to-government discussions are not routinely disclosed.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, what advice has been provided to government departments regarding the procurement and use of Chinese open weight AI models.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
I refer the Hon. Member to the answer given on 22 October to Question 28819.
The Government has also published the AI Cyber Security Code of Practice which is supported by additional guidance from the National Cyber Security Centre. The UK Code of Practice informed ETSI EN 304 223, the first global standard to define baseline cyber security requirements for AI models and systems, covering the entire lifecycle - design, development, deployment, maintenance, and decommissioning.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, how many Chinese open weight AI models used by Government Departments have been assessed by AISI.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
I refer the Hon Member to my previous answers to questions 28819 and 37340.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, whether she has made an estimate of the time saved per civil servant per day from generative AI tools.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
Artificial Intelligence is the single biggest transformational technology of our lifetime. It offers an unprecedented opportunity to redesign workflows, eliminate high-volume admin tasks, and deliver simpler, better, faster public services. Ethical and transparent deployment of AI with appropriate guardrails must be our overriding principle.
Although we have no formal estimate of time saved per civil servant per day from generative AI tools yet, there is already important work underway to capitalise on the opportunities AI presents and early indication on potential time savings, for example:
The Ministry of Justice’s in-house Justice AI Unit has rolled out its notetaking and case intelligence tool, Justice Transcribe, to 11,000 probation staff, which has been used in over 1.8 million meetings to date and is estimated to save around 450,000 hours of staff time annually.
“Consult”, a tool developed by i.AI to analyse consultation responses, saved around 8,000 hours of admin and over £250,000 on seven live consultations in 2025.
MHCLG’s AI augmented planning decisions support planners with policy checks, and report drafting, aiming to reduce processing time by 50%.
The Home Office has piloted AI to summarise asylum cases, with average time savings of 30%.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, whether the AI Security Institute evaluated Gemini 4 Carbon before its release on 28 October 2026.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
I refer the Hon. Member to the answers given on 2 November 2026 to questions 38845 and 38850.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, whether any of the £100 million for the AI Biosecurity Lab announced on 24 October 2026 is drawn from the £115 million that the Defence Investment Plan states DSIT will spend to strengthen defences against the risks of AI.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
Yes. The £100 million committed to the AI Biosecurity Lab forms part of the £115 million committed to strengthen the UK’s defences against risks arising from advanced AI, including the use of AI to improve biosecurity.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, what the budget of the AI Security Institute is in each financial year from 2026-27 to 2029-30 including the additional £120 million announced on 24 October 2026.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
The AI Security Institute (AISI) has been allocated £360 million from 2026-27 to 2029-30. This includes the additional £120 million over 2026-27 to 2029-30 announced at Budget 2026, on top of the £240 million set out in my response of 15 October 2026 to PQ 33991. The additional funding will strengthen the UK’s frontier AI evaluation and testing capability and support the development of technical standards to ensure risks are mitigated.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, pursuant to the Answer of 28 October 2026 to Question 36255, how does AISI decide which models count as 'frontier or near frontier'.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
The AI Security Institute (AISI) does not apply a fixed threshold. It uses the term for models where testing is most likely to change our understanding of emerging AI capabilities and risks. AISI evaluates this on a case-by-case basis and reviews its approach regularly as the technology develops. Through its trusted relationships with the world's leading AI developers, AISI continues to ensure its evaluations focus on the models that matter most.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, whether the AI Security Institute shares pre-deployment results with Canada's AI Safety Institute.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
The AI Security Institute (AISI) shares the findings of its pre-release evaluations with Canada’s AI Safety Institute in line with our individual agreements with AI developers. The renewed UK-Canada AI security agreement has also established regular exchanges on frontier AI capabilities, emerging risks, evaluation methodologies and best practice.
Asked by: Ben Spencer (Conservative - Runnymede and Weybridge)
Question to the Cabinet Office:
To ask the Minister for the Cabinet Office, how many persons and organisations affected by unsanctioned actions during AISI’s cyber evaluations of 25 to 28 July 2026 have been notified.
Answered by Kanishka Narayan - Minister of State (Cabinet Office) (Jointly with the Department for Business, Innovation, Science and Technology)
The AI Security Institute (AISI) notified all affected individuals and organisations where it was possible to do so. This included notification of the GitHub users the model interacted with.
I refer the Hon. Member to the technical report AISI published on its website on 4 August 2026. It sets out in detail how AISI detected, contained and responded to the incident, including a timeline of its notifications.