Phishing
AI phishing emails and what to do about them
Do not rely on bad grammar to catch a phishing email. The wording can now be polished, so the defence moves to the request itself and to the controls around it.
Last reviewed against NCSC (UK), UK Government (DSIT and Home Office) and Anthropic. Sources
The baseline
Phishing is already the main problem
Phishing reached 38% of UK businesses and 25% of charities in the 2025/2026 government survey, and the report called it by far the most prevalent type of breach or attack [4]. Among organisations that had a breach or attack, 69% said phishing was the most disruptive kind [4]. The figures quoted here do not say whether AI was used, so they describe the size of the problem, not how much AI adds to it.
What AI adds
Three things to assume
Fluent, tailored text
The NCSC judges that AI will help attackers with some of the steps in an intrusion [1]. Plan on lures that read well and can be adapted to a role or a company.
More attempts
Anthropic said AI has lowered the barriers to sophisticated cybercrime, so people with few technical skills can attempt more [5]. Expect more attempts, not only better ones.
Targeting from public information
The NCSC advises limiting public information that could be used for spear phishing, and keeping an eye on what partners and suppliers reveal about you [3].
Controls
What still works
Apply anti-spoofing controls (DMARC, SPF and DKIM) to your domains, and keep email filtering on for all users [3]. Use multi-factor authentication so that a stolen password alone is not enough [3]. Consider password managers or single sign-on, which can help users notice fake sites [3].
Add these measures too:
Why wording is no longer the test
Staff who have been taught to look for typing errors will pass a clean message. Teach them to ask what the message wants them to do, and whether that action is one they would normally take on the strength of an email.
Keep reading
Read next
Voice scams
What UK research says about staff recognising voice-based scams, how fake audio is reported by security leaders and how to confirm a caller safely.
Read the guide
Fake applicants
Anthropic reported North Korean operatives using AI to win remote technical jobs under false identities. What UK employers can take from the case.
Read the guide
Prompt injection
What prompt injection is, why the NCSC says it may never be fully fixed, and the practical limits that reduce its impact on your AI tools.
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.