Your AI tools
Securing the AI tools you already use
Staff adopt AI tools faster than security practices follow. Start with what each tool can read, what it can do, and who can see its logs.
Last reviewed against NCSC (UK), NCSC New Zealand and UK Government (DSIT and Home Office). Sources
The gap
Adoption ahead of controls
In the 2025/2026 survey, 31% of businesses and 25% of charities were using AI, adopting it or considering it. Of those, 24% of businesses and 27% of charities had security practices for AI-related risks [5]. So most organisations using AI have not yet put dedicated controls in place.
The checklist
Questions to ask of every tool
-
1
What data can it read, and is any of it untrusted, such as inbound email or web pages? [2] -
2
What tools, APIs or accounts can it use, and with whose privileges? [2] -
3
Could a deterministic control outside the LLM limit those actions? [2] -
4
Are inputs, outputs, tool use and API calls logged? [2] -
5
How does the vendor handle your source code and sensitive data? [4] -
6
Is use restricted to a test environment or sandbox where it can be? [4]
Risk names
The OWASP list in brief
OWASP's 2025 list includes prompt injection, sensitive information disclosure, supply chain, data and model poisoning, improper output handling, excessive agency, system prompt leakage, vector and embedding weaknesses, misinformation and unbounded consumption [8]. Excessive agency is the one to check first, because it is about what an LLM is allowed to do.
Stop and decide
If a tool must read untrusted content, it should not hold privileged access. Remove one or the other [2].
Governance
Who owns it
Give each AI tool an owner and put it on your asset inventory, in the same way you would any other system that touches company data. The NCSC's first step for exposure is knowing what you have [4], and AI tools adopted by individual teams are easy to miss.
Write down, for each tool, what it is allowed to read, what it is allowed to do and who reviews its logs. A short record like this makes it possible to answer a board or auditor quickly when they ask how AI use is controlled. See prompt injection for the design measures.
Keep reading
Read next
Board briefing
What the Five Eyes cyber agencies ask of leaders on AI risk, and how to turn it into a short, sourced board briefing for a UK organisation.
Read the guide
UK statistics
The UK figures on breaches, phishing and AI adoption from the 2025/2026 Cyber Security Breaches Survey, and what they do and do not tell us about AI.
Read the guide
Am I exposed?
Check whether your people, internet-facing systems and AI tools are exposed to AI-assisted attacks, with sourced pointers for UK organisations.
Read the guideHow 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?
AI makes it quicker to find and use weak points, so it helps to find them first. A penetration test shows which of your internet-facing systems, staff processes and AI tools an attacker could actually use.