Latest
Breaches survey reports AI adoption and AI security practice[10]
The government's survey found 31% of businesses using, adopting or considering AI and about a quarter of adopters with AI security practices.
Overview
AI security: the controls that matter for UK organisations
AI security is two jobs: protecting the AI tools you run, and using AI safely in defence. UK guidance says some weaknesses cannot be patched, so controls on what AI tools can reach and do come first.
Last reviewed against NCSC, Department for Science, Innovation and Technology and OWASP GenAI Security Project. Sources
The latest
AI security controls: what has happened so far
Newest first. Each entry links to its source.
NCSC says prompt injection is not SQL injection[1]
The NCSC argued that prompt injection may never be fully mitigated and that the aim should be to reduce its likelihood and impact.
NCSC assessment of AI and cyber threat to 2027[4]
The NCSC assessed how AI is likely to change the threat to the UK, including a larger attack surface as AI use spreads.
If your staff use chatbots, assistants or AI agents, or you are buying products that now include them, this site sets out what the NCSC, the UK government and OWASP advise. It stays with what their publications say.
The explainer
Two jobs, one subject
Securing AI systems
The tools you build or buy can be manipulated, can leak data and can be handed too much access. The NCSC names prompt injection and data poisoning among the main weaknesses of LLMs [2]. Start with what AI security is.
Using AI in defence
The NCSC encourages organisations to use AI to strengthen their defences, and notes that suppliers already use it to find and fix vulnerabilities [3]. Outputs still need checking: see AI in cyber defence.
Governing the lot
Someone must own it. The NCSC asks who is accountable for AI security, what the worst failure would look like and how well the supply chain is understood [2]. The controls guide turns that into a working set.
What changed
Why it matters now
Adoption is under way. The 2025/2026 government survey found 31% of businesses using, adopting or considering AI [10]. Among businesses using or adopting it, 24% had cyber security practices to manage AI-related risks [10].
The NCSC's assessment to 2027 expects attackers to gain efficiency from AI in parts of their work, and says wider AI use across the UK's technology base enlarges the attack surface [4]. In December 2025 it added that prompt injection may never be completely mitigated, so effort belongs in lowering its likelihood and its impact [1].
The practical consequence
Some uses cannot absorb the residual exposure. For those, the NCSC says an LLM may not belong in the design at all [1].
Where to start
First things to do
Begin with the reach of each AI tool, meaning the data it sees and the actions it takes, not with the attacker's wording. The checker below takes three questions, and the practical plan sets out the rest. The risks page gives the evidence behind each control, and the questions page has short answers.
What to do
Four things to do first
-
1
List every AI tool and what it can reach
IT and security leads
-
2
Cut each tool's privileges to the minimum
Development and security teams
-
3
Put a person in front of high-risk actions
Product owners and system owners
-
4
Log inputs, outputs and tool use
Security operations
There are seven steps in all, with who does each one. See every step
Am I affected?
Answer a few questions
Quick answers
Questions people ask
What is AI security?
It means protecting the AI systems you build or use, and using AI safely to support defence[3]. The government's code of practice treats AI security as a subset of cyber security[6].
What are the main AI security risks?
The NCSC points to prompt injection, data poisoning, hallucinations and bias[2]. OWASP's 2025 list ranks prompt injection first and sensitive information disclosure second[8].
About this site
Primary sources, dated
The CyPro Consulting Ltd team maintains this site. Every version, date and figure comes from 5 named sources, including NCSC, Department for Science, Innovation and Technology, OWASP GenAI Security Project, 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
Risks
The AI security risks UK organisations face, from prompt injection to data exposure and over-broad access, with NCSC and OWASP evidence.
Read the guide
Plan
A practical AI security plan for UK organisations: inventory, least privilege, human approval, logging and ownership, drawn from NCSC advice.
Read the guide
Questions
Sourced answers on AI security: what it is, whether prompt injection can be fixed, shadow AI, the UK code of practice and what to do first.
Read the guideGo deeper
In depth Controls 
In depth Code of practice 
In depth Prompt injection 
In depth AI agents 
In depth Shadow AI 
In depth AI for defence 
In depth Suppliers 
Question What it is 
In depth OWASP list How CyPro can help
Need a hand planning for this?
Understanding the change is one job. Working out what it means for your systems, your suppliers and your roadmap is several more. CyPro's consultants can take any of them on.
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
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
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
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
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
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
Plan ahead
Want a plan for your organisation?
AI security starts with knowing which tools are in use, who owns them and what they can reach. CyPro can help set the governance, accountability and controls around that.