← Back to blog
AI cybersecurityAI policy

Faster attacks, familiar basics: AI cybersecurity for SMEs starts with your own AI policy

AI cybersecurity for SMEs: how criminals use AI, why the basics still work and how an AI policy keeps your team's own AI use safe.

16 Nov 2026·8 min read

Attackers now use AI in every phase of a cyberattack, and they do not need the most advanced models to do it, the Dutch National Cyber Security Centre (NCSC) wrote in its Cyber Security Assessment Netherlands 2026. Yet the police and the Public Prosecution Service (Openbaar Ministerie) conclude that AI does not fundamentally change cybercrime; it makes it faster, larger and more effective. For an SME that means two things. The basic measures you already know still do most of the work. And the new risk sits closer to home than most people think: the AI tools your own employees already use. AI cybersecurity for SMEs therefore starts with an AI policy that says which tools are allowed, what data may go in, and how people check what comes out.

How are criminals using AI against companies?

Mostly to do the same things faster and at a larger scale. The police and prosecution service name phishing, developing malware and searching for vulnerabilities as the main uses. The NCSC adds that AI helps attackers write more convincing phishing emails, which removes one of the old warning signs: the clumsy sentence or the odd spelling mistake.

The Dutch Data Protection Authority (Autoriteit Persoonsgegevens, AP) sees the effect in its own figures. The number of reported account takeovers in the Netherlands rose from 607 in 2024 to 1,742 in 2025, almost three times as many, according to the AP’s data breach report published in July 2026. The AP also points to ready-made phishing kits that let attackers start without much technical knowledge.

An opinion piece in AG Connect recently called this a reason for a new security mindset. We agree with the mindset, but would put the emphasis differently. For most SMEs the change is not a new kind of threat. It is that the old weak spots get found and used more quickly.

Why are SMEs more exposed than large companies?

Because the gap is in capacity, not in awareness. The National Coordinator for Security and Counterterrorism (NCTV) writes that the SME sector is extra vulnerable and that many organisations do not have their basic measures in order. The NCSC is even more direct: small organisations and SMEs cannot raise their resilience on their own.

Three forces come together in a company of 20 to 250 people.

Speed. When attackers automate the search for weak spots, an unpatched server or a reused password gets found sooner. The time between a known vulnerability and an attempt to use it shrinks.

Credibility. A payment request that sounds exactly like your supplier, or a voice message that sounds like your director, takes more than a gut feeling to spot. The checks need to be in the process, not in someone’s instinct.

Your own AI use. This is the force companies control most directly, and the one that gets the least attention. We come back to it below, because it is where an SME can make the biggest difference with the least money.

What is shadow AI, and why does it matter for security?

Shadow AI (schaduw-AI) is the use of AI tools at work without the employer knowing about it or having approved it. It is common. In the 2024 Work Trend Index of Microsoft and LinkedIn, a survey of 31,000 knowledge workers in 31 countries including the Netherlands, 78 percent of AI users said they brought their own AI tools to work. At small and medium-sized companies that was 80 percent. More than half (52 percent) were reluctant to admit using AI for their most important tasks.

In practice this means your employees probably already work with AI, often with good intentions: they want to write faster, summarise a long document or tidy up a spreadsheet. The risk is in what they paste in. The AP received several reports of data breaches where employees entered personal data into a chatbot, often on their own initiative and against company agreements. One example was a GP practice that entered patients’ medical data; another was a telecom company that entered a file with customer addresses. The AP notes that most companies behind chatbots store everything that is entered.

That is why we see an AI policy as a security measure, not a paperwork exercise. Banning AI rarely works when tools are a browser tab away. Making the safe route the easy route does.

How do you write an AI policy that keeps your company safe?

There is no public SME case we can point to with measured results, so here is a step-by-step plan built on the guidance of the NCSC on using AI safely and the AP’s advice. Each step is an action plus the reason for it.

Step 1: Find out what is already being used

Ask your teams which AI tools they use, including those outside the official IT environment. Do it as an open question, not an audit. The reason: you cannot set sensible rules for tools you do not know about, and people will only tell you if they do not expect a reprimand.

Step 2: Choose the tools you make available

Pick one or two business versions of AI tools and arrange that prompts and input are not stored or used for training. The NCSC explicitly mentions such contract agreements. The reason: when a good, approved tool is available, the pull towards a private account disappears.

Step 3: Agree what data may go in

Write down in plain language which information may be used as input and which may not, such as personal data, client files or financial details. The AP advises exactly this. The reason: most chatbot data breaches start with one well-meant copy-paste.

Step 4: Build checks into the process

Agree verification rules for payments and changes to bank details, for example always calling back on a known number. The NCSC and the police gave the same advice for voice clones and fake invoices. The reason: when a message can sound perfectly real, the check has to be a rule, not a feeling.

Step 5: Train people and review once a year

Explain the policy in a short session with real examples from your own work, and add an AI check to your yearly risk analysis, as the NCSC suggests. The reason: this also meets a legal duty. Under Article 4 of the EU AI Act, companies that use AI must take measures for the AI literacy (AI-geletterdheid) of their staff. According to the European Commission’s Q&A, this has applied since 2 February 2025, and the amendments of the Digital Omnibus that entered into force in mid-July 2026 keep it an obligation for providers and deployers. No certificate is required; an internal record of training is enough. This is not legal advice.

Can AI also help protect an SME?

Yes, and the NCSC calls for investment in AI for defensive capabilities as well. For a smaller company the realistic route is usually through the tools you already pay for: email filtering, endpoint protection and the security features of your cloud suite increasingly use AI to recognise unusual patterns. The point is less about buying an “AI security” product and more about making sure those features are switched on and someone looks at the alerts.

The NCSC also reminds us that the basics do most of the work: good patch management, a small attack surface, and logging and monitoring. AI does not replace those basics. It makes them more urgent, because attackers find the gaps faster.

From loose tools to a way of working

The common thread is that safety comes from how the work is organised, not from one product. A company where everyone uses their own chatbot with their own habits has many small, invisible risks. A company where AI has a fixed place in the process, with an approved tool, clear rules on data and a check before money moves, has fewer and knows where they are. That is also the company that gets the most value out of AI.

Not every SME has someone in-house to set this up. That is where outside expertise helps: someone who can map the current use, configure the tools and train the team in a few sessions. Read more about how we help SMEs put AI to work, or look at the AI services from experts, including policy and training.

Frequently asked questions

Does my SME need an AI policy?

If your employees use AI tools at work, a short AI policy is the simplest way to keep that safe. It tells people which tools they may use, what data may go in and how to check the output. It also helps you show that you take AI literacy seriously, which the EU AI Act asks of companies that use AI.

What is shadow AI?

Shadow AI is the use of AI tools at work without the employer knowing or approving it, for example a personal ChatGPT account used for client emails. It is usually well meant. The risk is that confidential or personal data ends up with a provider that stores it.

Is AI making cyberattacks on small businesses more likely?

According to the Dutch police and prosecution service, AI does not fundamentally change cybercrime, but it does increase its speed, scale and effectiveness. The NCSC says attackers use AI in all phases of an attack, including more convincing phishing. Weak spots are therefore found sooner, which makes basic measures more important.

What should an AI policy for employees include?

At minimum: which AI tools are approved, which data may and may not be entered, how output is checked before it is used, and who to ask with questions. Add verification rules for payments and a yearly review. Keep it short enough that people actually read it.

Your next step

Start with one question to your team this week: which AI tools do you use, and for what? The answers tell you where your policy needs to begin. For more background, see how posting a question works for companies. If you want help setting up approved tools, rules and a short training, post your question on Baionic: within one working day you receive proposals from one to three matching experts, with their approach, price and planning side by side.

Ready to take the next step?

Post your project for free and receive proposals within one business day.

Post a project

More reading