What do AI-powered cyber attacks mean for a small business?

AI-powered cyber attacks mean the gap between a security flaw being found and being used against you has shrunk from weeks to hours. For a small business, the practical answer is to confirm your website software is patched, review which outside apps can sign in to your business accounts, and make sure every tool you connect, including any AI tool, can reach only what its job requires. None of these steps is new, but artificial intelligence (AI) has made leaving them undone far more costly.

Why this is in the news

Two linked stories made October 2026 the month some security writers have started calling 'Black October'.

On 21 October the PHP project released emergency fixes for two critical flaws in PHP (Hypertext Preprocessor), the programming language behind an estimated 70% of websites, including most WordPress sites. Researchers had found the flaws, nicknamed 'PHP2Shell', with the help of AI models. They allow an attacker to run code on a web server with a single request and no login. Amazon's threat intelligence team reported seeing the first attempts on 17 October, four days before the fixes were public. On 26 October the US Cybersecurity and Infrastructure Security Agency (CISA) added both flaws to its list of vulnerabilities being actively exploited, and several governments took some of their sites offline as a precaution.

The same day as the PHP fixes, Microsoft reported that a China state-aligned group it tracks as Storm-0940 had compromised a cloud backup service run by ConnectWise, an information technology (IT) management provider. Using stolen app credentials, the group set an AI agent framework called ARTEX to work. It attempted sign-ins to more than 2,250 Microsoft 365 business accounts at over 2,900 attempts a minute, a pace no human team could match. Microsoft said most attempts failed, but at some organisations the attackers gained privileged access in under three hours.

What AI-powered cyber attacks change for a small business

The lesson is speed rather than a new kind of attack. The ConnectWise breach used permissions that customers had already granted to a trusted supplier. PHP2Shell targeted software that many businesses run without realising it. AI simply made finding and using those openings faster and cheaper, which puts smaller firms within easy reach of attackers who once had to choose their targets with care.

This matters because a company with 2 to 50 staff rarely has someone watching security alerts around the clock. If an attacker can move from first contact to administrator access in under three hours, the protection has to be in place before the attack, not after it.

Four checks to make this week

Confirm your website is patched. Ask your web host or developer whether your site runs PHP and, if it does, whether it is on version 8.5.12, 8.4.27, 8.3.36 or 8.2.35 or later. Get the answer in writing, with the date the update was applied.

Review the apps connected to your business accounts. In Microsoft 365 or Google Workspace, list every outside app that has been granted access. Remove anything no longer used, and ask your IT provider whether any connection came through a supplier affected this month.

Turn on stronger sign-in protection. Multi-factor authentication (MFA), ideally with a hardware key or an authenticator app, blocks most automated password attempts. It is the cheapest defence against the kind of high-speed sign-in attack Microsoft described.

Limit what every tool can reach. Each connected service, including AI agents for small business work, should have the narrowest access its job needs. We set out how to do this in what to check before giving an AI agent access to your systems.

Should this put you off AI automation?

We do not think so, but it should change how you choose and set up AI tools. The risk this month came from broad, long-lived permissions and unpatched software, and a well-built AI tool is designed to avoid both.

When we build a custom AI tool, such as an AI receptionist or an automated booking system, we start from the same rules: it connects only to the systems it needs, it uses credentials that can be withdrawn quickly, it keeps a record of what it did, and a person approves anything that cannot be undone. Those choices matter more than which AI model sits underneath. If you are new to the idea, our guide to what an AI agent is and what it can do is a good starting point, and our solutions page shows the tools businesses deploy today.

AI is now on both sides of cyber security. Attackers use it to move faster, and businesses can use well-designed AI automation to handle routine work without opening new doors. The difference lies in how carefully the tools are set up.

If you would like to talk through which AI tools fit your business and how to deploy them safely, book a call with us.

Written by the AI Automators team.

← All posts