Your Privacy
English

Bodies with power

AI Security Institute (AISI)

Type
Government body
Place
United Kingdom
Last checked
2026-09-22
Next check due
2027-03-21

Ways to reach them

  • Data protection / privacy enquiries (DSIT DPO, covers AISI) · Email address
    Verified · 2026-09-22

    Privacy and data-protection questions only. Published on AISI's privacy policy and monitored by DSIT's Data Protection Officer.

    VERIFIED at https://www.aisi.gov.uk/privacy-policy. This is the ONLY email address AISI publishes anywhere on its site.

  • AISI contact page · Web form
    Unverified

    Linked from the site navigation.

    UNVERIFIED — the page exists in AISI's navigation but could not be retrieved (permission/provenance failure on fetch). Its contents could not be confirmed, so do not assume an email is published there.

  • Research collaboration and careers · Web form
    Unverified

    AISI recruits researchers and runs grant programmes; this is the realistic route in for technical people.

    UNVERIFIED as fetched; the careers path is linked from the homepage, which WAS fetched.

  • Public complaint or incident route · Complaint route
    none
    Verification expired · 2026-09-22

    This route is retained as reference only. Its verification is missing, more than 180 days old, or the contact window is not currently open; do not rely on it until it is re-checked.

    n/a

    VERIFIED ABSENCE — AISI is a research body with no complaints function, no regulatory remit and no public reporting channel.

  • Homepage · Homepage
    Unchecked

What it does

RENAMED: was the 'AI Safety Institute'; renamed the AI Security Institute in February 2025 (announced by written ministerial statement, HLWS454, 24 February 2025), with the focus narrowed toward security, cyber, chemical/biological and criminal-misuse risks and away from bias and free-speech concerns. It is a DIRECTORATE OF DSIT, not an independent agency. Conducts pre-deployment and post-deployment testing of frontier models by voluntary agreement with developers, builds evaluation infrastructure, publishes a Frontier AI Trends Report, and advises government. It has NO statutory powers: it cannot compel access to a model, cannot fine, cannot require testing, and cannot stop a release.

Its powers

research

Honest assessment

AISI publishes essentially no way for a member of the public to reach it — one data-protection address, maintained by its parent department. It is deliberately a research organisation, not a regulator, and there is nothing to complain to it about. What DOES reach AISI: publishing technical evaluation work it cites, applying to its research grant calls or open roles, and responding to DSIT consultations that shape its remit. The February 2025 rename was substantive, not cosmetic — 'safety' framing now lands less well than security, misuse and national-resilience framing. Because AISI is a DSIT directorate, parliamentary scrutiny of it happens through the Commons Science, Innovation and Technology Committee, which is where public evidence about AISI actually gets read.

Notes

Frontier/catastrophic risk IS squarely in scope — chem/bio, cyber and autonomy evaluations — making it the UK body most focused on severe risk, but also the one least reachable.

Sources

Something wrong here?