Latest
Cyber Security Breaches Survey 2025/2026 published[6]
It found 43% of businesses reported a breach or attack, with phishing the most common, and that few AI-using organisations had practices to manage AI risk.
Overview
AI powered cyber attacks: what they are and how to defend
Attackers use AI to make scams more believable and intrusions faster, and AI tools are a target too. The NCSC expects AI to sharpen existing methods, so tested basics still carry most of the defence.
Last reviewed against NCSC (UK), Anthropic and UK Government (DSIT and Home Office). Sources
At a glance
- 43%
- of UK businesses reported a breach or attack in the last 12 months[6]
- 38%
- of UK businesses were hit by phishing, the most common type[6]
- 24%
- of businesses using or considering AI had practices to manage its risks[6]
- 80-90%
- of one reported espionage campaign was carried out by AI, per Anthropic[5]
The latest
AI powered cyber attacks: what has happened so far
Newest first. Each entry links to its source.
NCSC: prompt injection is not SQL injection[2]
The NCSC said prompt injection probably cannot be fully prevented and advised reducing its likelihood and impact through design.
Anthropic publishes details of the campaign[5]
It reported about thirty targets, a small number of successes and AI doing 80-90% of the work, with humans at a few decision points.
The full timelineThis site follows one question: what does AI change about the cyber threat to a UK organisation, and what should a decision maker do about it? It draws on the NCSC, the government's breaches survey, a vendor's own account of a reported espionage campaign and one well-documented fraud.
At a glance
Three kinds of AI powered attack
The phrase covers three separate things, and the right response differs for each.
Better scams and impersonation
AI helps produce messages, voices and video that look and sound like people you know. An employee at Arup in Hong Kong sent HK$200m (about £20m) to criminals after a video call in which synthetic voices and images stood in for senior officers [8]. See AI phishing and impersonation.
Faster, more scalable intrusion
In the NCSC's assessment, AI is set to keep improving parts of how intrusions are run, in both effect and efficiency [1]. Anthropic reported a campaign in which its coding tool was manipulated into doing 80-90% of the work against about thirty targets [5]. See AI-assisted intrusion.
Attacks on AI tools
Prompt injection is ranked first in OWASP's 2025 list of LLM application risks [9]. The NCSC says it probably cannot be fully prevented, so design has to limit the damage [2]. See prompt injection defence.
The UK picture
What the survey shows
Over the last 12 months, according to the government's 2025/2026 survey, a breach or attack hit 43% of UK businesses, and 28% of charities [6]. Phishing led the list at 38% of businesses [6], and 69% of those affected named it the most disruptive type [6].
The survey does not blame AI for those figures. The AI finding is about readiness: among businesses that use AI or are moving towards it, 24% had practices in place to manage its risks [6].
Where to start
Pick the one process where a convincing request could move money or grant access, and add a second check through a separate channel. The Arup case turned on a video call that looked real [8].
Next steps
Where to go from here
Use the exposure checklist to see which of the three areas applies to you, then the defence plan for ordered actions. Questions people ask most are on the questions page.
What to do
Four things to do first
-
1
Add a second check to payments and access resets
Finance, HR and service desk teams
-
2
Make reporting a suspicious message easy
Security leads and managers
-
3
Keep patching and configuration basics current
IT and operations teams
-
4
Reduce what an AI tool can do when it reads outside content
Developers and system owners
There are seven steps in all, with who does each one. See every step
Am I affected?
Answer a few questions
Check if you are affected
Answer three questions about your payment checks, your email defences and your AI tools. Nothing you answer leaves this page.
Could a payment, a bank detail change or an access reset go ahead because someone on a call, video meeting or message appeared to be a senior colleague?
Quick answers
Questions people ask
What is an AI powered cyber attack?
It is an attack where criminals or state-linked actors use AI to scale or improve their methods, such as phishing, impersonation and intrusion. The NCSC expects AI to keep making parts of an intrusion quicker and more effective[1]. Attacks on AI tools, such as prompt injection, are a related but separate category[2].
How common are cyber attacks on UK organisations?
The 2025/2026 government survey counted 43% of businesses and 28% of charities as having had a breach or attack in the past year[6]. The survey does not attribute those figures to AI[6].
How did the Arup deepfake fraud work?
An employee was tricked into transferring HK$200m (about £20m) after a video call in which synthetic voices and images impersonated senior officers[8]. The company said its own systems had not been breached[8].
Has AI really carried out a cyber attack by itself?
Anthropic reported a campaign it assessed as Chinese state-sponsored in which AI did 80-90% of the work, with humans making perhaps four to six critical decisions per campaign[5]. The company describes it as the first documented case of a large-scale attack executed without substantial human intervention[5].
About this site
Primary sources, dated
The CyPro Consulting Ltd team maintains this site. Every version, date and figure comes from 7 named sources, including NCSC (UK), Anthropic, UK Government (DSIT and Home Office), and links to the one it came from. Last checked 4 November 2026.
Who we are and how pages are kept currentStart here
The guides, in the order to read them
Am I exposed?
A practical checklist to see whether your payment processes, inboxes, systems and AI tools are open to AI powered attacks, for UK organisations.
Read the guide
How to defend
An ordered plan for UK organisations to defend against AI powered attacks: payment checks, phishing, patching, AI tool limits, logging and the board.
Read the guide
Timeline
A dated, sourced timeline of AI related cyber attack cases and UK guidance, from the Arup deepfake fraud in 2024 to the 2026 breaches survey.
Read the guide
Questions
Sourced answers on AI powered cyber attacks: what they are, how deepfake fraud works, prompt injection, UK figures and what organisations should do.
Read the guideGo deeper
In depth AI phishing 
In depth Deepfake fraud 
In depth AI intrusion 
In depth Prompt injection 
In depth AI tools at work 
Question Are they new? 
Question Can it be fixed? How CyPro can help
Need a hand with any of this?
Patching is one job. Keeping a vulnerable system safe until it is patched, and proving it, is several more. CyPro's consultants, testers and SOC analysts can take any of them off your plate.
Vulnerability scanning
Find every internet-facing system you run, including the forgotten ones, and see which are still on a vulnerable build.
See the scanning service
Hardening and segmentation
Our consultants put the controls in place that limit the damage when something is missed: firewall rules, network isolation and tighter admin access.
More on cypro.co.uk
24/7 security monitoring
Our SOC analysts watch your systems day and night for the signs of an attack in progress: unexpected processes, new outbound connections, logins that do not fit.
More on cypro.co.uk
Penetration testing
A focused test of your systems and the controls around them, so you know they hold before an attacker finds out for you.
More on cypro.co.uk
Incident response
If you may already be compromised, our responders contain it, preserve the evidence, find out what was touched and get you running again safely.
More on cypro.co.uk
Fractional CISO for the board
A senior security lead who advises the board on what to fix first, what to accept and what to tell customers, and owns the decision record.
More on cypro.co.uk
Know your exposure
Not sure you are protected?
Many of the risks here sit in how an organisation uses AI: what tools staff run, what they can reach and who owns the risk. AI governance work sets that out so controls match the tools in use.