The BaseOfficialGOV.UK

Britain steps up national defences against growing AI risks

In a press release, the UK government says it will use frontier AI to strengthen cyber defences, consider tougher rules on serious AI incidents and loss of control, and prepare for AI-agent incidents; its AI Security Institute is evaluating the newest models. AI Minister Kanishka Narayan tells AI companies not to release systems whose risks they cannot adequately control.

Open the source ↗Top story of the week

Why it matters

The same thing, for different readers.

What it means for you

The UK says it will use AI to protect public services such as hospitals from AI-enabled attacks, and is planning for incidents where AI agents act on their own.

For lawmakers

The UK is exploring extending its Cyber Security and Resilience Bill to cover AI loss of control and the reporting of serious AI incidents and near misses.

For researchers

The release names AI verification, checking claims about how models were developed and tested, as a field the UK will support, with details promised in coming months.

The question it raises

A minister telling companies not to release what they cannot control leaves open who judges when control is adequate: the company or the state.

In their own words

Every quotation was found, word for word, in the text we saved from the source.

From the source’s own text.

“And to the companies developing the world’s most powerful AI systems, my message is: Listen when your scientists warn of serious dangers. Investigate unexpected behaviour. And if you cannot adequately control the risks, do not release the system.”
Kanishka Narayan, AI MinisterThe AI Minister's statement in the government press release, issued as Narayan spoke at Imperial College London's Base One incubator.
Matches the source exactly
“We will use every power at our disposal. We will legislate wherever necessary. And we won’t look to law alone.”
Kanishka Narayan, AI MinisterThe AI Minister's statement in the press release, right after saying Britain should be one of the hardest countries in the world to attack using AI.
Matches the source exactly
“the government will explore strengthening parts of the Cyber Security and Resilience Bill and the wider regime, to ensure the UK’s essential services such as hospitals and utilities are better protected from emerging AI related risk. This includes AI loss of control risk, where autonomous AI systems evade human oversight.”
Uk GovernmentThe press release's own text on possible new legal protections; the release is issued jointly by the Cabinet Office, AI Security Institute, NCSC and one further department.
Matches the source exactly
“A major incident involving AI agents could play out quicker and wider than other major incidents that governments are designed to respond to, and rising to this challenge will require the development of new expertise and capabilities.”
Uk GovernmentThe press release on joint work by the National Cyber Security Centre and the AI Security Institute on responding to incidents involving AI agents.
Matches the source exactly

The record

Questions
Who can demand evidence, delay a release or stop an unsafe deployment? · How is AI used in attacks, defence and war? · Can AI systems act beyond what people intended or allowed?
People
Kanishka Narayan (Speaks, AI Minister) · Liz Lloyd (Speaks, Cyber Minister)
Also covered by
The Guardian · BBC News
What we read
We read the whole source.
Found by
The Radar, run of October 12, 2026, via the-guardian-ai (tier 2: a major outlet’s own reporting)
Cite this
GOV.UK, “Britain steps up national defences against growing AI risks”, October 12, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/gov-uk-2026-10-12-britain-steps-up-ai-defences
Our notesTriage main item was the Guardian story; rebuilt on the official GOV.UK press release (triage 'also' c6a72e78a1ce) as the primary source, with the Guardian and BBC as other coverage. url/outlet/title/published are therefore GOV.UK's, not the Guardian's. Guardian and BBC were read but not quoted: the Guardian says Narayan 'will say' Britain cannot rely entirely on tech companies to police their own models (not in the GOV.UK text); the BBC reports a press-conference remark that a breach is 'likely, if not inevitable' and a reference to a rogue OpenAI agent hacking an Australian government website; both are worth a separate check if we want them. The Guardian also covers Geoffrey Irving, Satya Nadella and an Anthropic test in which Claude contacted police (not used here; that part would be a conflict item). The release's 'From' line names the 'Department for Digital, Culture, Media and Sport' as written; our text does not repeat the department name. The release says the Prime Minister spoke at the UN General Assembly in September but does not name the Prime Minister.