AI tools
How to secure the AI tools your organisation uses
Attackers can also target the AI tools you adopt. Start with an inventory, limit what each tool can reach and do, and ask suppliers how they secure theirs.
Last reviewed against NCSC (UK), UK Government (DSIT and Home Office) and Anthropic. Sources
Why
Adoption is ahead of controls
Government survey data for 2025/2026 shows 31% of businesses using, taking up or weighing AI, and only 24% of that group had security practices aimed at AI risks [4]. Most organisations in the group therefore lack specific controls.
The standards
Where to look
A UK government code of practice lays out baseline principles for securing AI systems, and the government planned to put it to ETSI as the basis for a global standard [6]. The NCSC's guidance for AI adopters stresses that security must be a core requirement across the AI lifecycle and that leaders should have a response plan ready [9].
Steps
A practical order
-
1
List every AI tool in use, including those staff adopted themselves. -
2
Record what each can read and what actions it can take. -
3
Reduce its privileges where it processes untrusted content [2]. -
4
Log inputs, outputs and tool calls [2]. -
5
Check that data used with AI meets legal requirements and good practice [9]. -
6
Ask suppliers about their own security and their data, model and software supply chains [9].
Questions
What to ask a supplier
What can the tool access?
Ask which data, accounts and systems it reaches and whether you can narrow that.
What happens if it is manipulated?
Ask what safeguards exist that do not depend on the model behaving, in line with the NCSC's advice on secure design [2].
What is logged?
Ask what you can see after an incident.
Responsibility
The NCSC says the burden of using AI safely should not fall on individual users, and that developers should take responsibility for their customers' security outcomes [9]. Hold suppliers to that.
Keep reading
Read next
Timeline
A dated, sourced timeline of UK guidance, surveys and company reports on AI-assisted social engineering, from February 2024 to April 2026.
Read the guide
Questions
Sourced answers on AI phishing, voice scams, prompt injection and what UK organisations should do, from the NCSC, government and published research.
Read the guide
Am I exposed?
A checklist for UK organisations: which requests, teams and AI tools could be manipulated by AI-assisted phishing, impersonation or prompt injection.
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?
Defending against AI-assisted social engineering is a matter of decisions: who may approve what, how requests are confirmed and what AI tools may do. A virtual CISO helps leaders set those rules and have them tested.