AI cyberattacks on critical infrastructure are coming, and they will not stop at power grids and water treatment plants. Richard Horne, head of the U.K.’s National Cyber Security Centre, warned that the U.K. will likely face AI cyberattacks on critical infrastructure by 2028 as criminals and states, including China and Russia, increasingly use AI in their cyber activity. The scale of these attacks will be unprecedented, driven by AI-powered malware that can discover and exploit vulnerabilities that have been tolerated for far too long.
For small and mid-sized businesses, you need to know if your operation counts as critical infrastructure. The question is whether the tools, vendors, and systems you rely on are ready for attackers armed with AI.
What does AI-powered malware mean for your business?
Horne’s warning centers on AI-powered malware for ransomware attacks. The technology will allow vulnerabilities to be discovered and exploited at a scale that was once unimaginable. The U.K. National Cyber Security Centre’s report noted that Chinese cyber activity is the most sophisticated and sustained cyber challenge facing the U.K., while both Russia and China will routinely use AI to achieve their geopolitical goals. Russia is intentionally conducting cyber campaigns which fall below the threshold of armed conflict, and Iran is carrying out attacks against the U.K. and targeting individuals who are a threat to the Iranian regime.
Picture a Monday morning when your accounting software will not open, your file server is locked, and a ransom note appears. The attacker did not guess a weak password or trick someone into clicking a link. Instead, an AI agent scanned your vendor’s public-facing systems, identified an obscure configuration flaw, and built a working exploit while you slept. The attack moved faster than your patch schedule, and it hit dozens of other firms the same night.
The report made clear that AI is more likely to be used to enhance and scale existing tactics and techniques rather than find brand new vulnerabilities. That means the gaps you have lived with because they seemed too small to matter will become the entry points attackers hit first.
How will attackers use AI against small businesses?
State actors and organized criminals will use AI to achieve efficiency and scale. The use of AI by states and criminals is highly likely to increase the scale and efficiency of hostile operations. Governments and corporations worked on urgent security updates after the warning was published, a sign that the threat is immediate, not theoretical.
For your business, the exposure comes through your vendors, your software, and your own internet-facing systems. If your firewall, VPN, or remote-access tool has a known vulnerability and you have not patched it, assume an AI agent has already cataloged it. If your managed service provider uses a platform with weak authentication, that platform is now a target. If your employees use AI tools without oversight, those tools are generating data that can be intercepted, manipulated, or exfiltrated.
Treat every unpatched system as already found. Treat every AI tool your team uses as a potential leak. Treat every vendor connection as a door that needs a lock.
What should SMBs do before AI-powered attacks arrive?
Start with an inventory of every AI tool in use across your organization. Include the chatbots, the browser extensions, the sales automation platforms, and the coding assistants. Ask who approved each one, what data it can access, and where that data goes when the tool processes it. If you cannot answer those questions in a day, you are not ready.
Next, audit your internet-facing systems. List every firewall, VPN, remote desktop gateway, and web application your business exposes to the internet. Confirm that each one is patched to the latest stable version. If a vendor has issued an update in the past six months, apply it. If a system cannot be patched because it is too old or unsupported, replace it or remove it from the internet.
Review your vendor risk program. Identify every provider that handles your data, connects to your network, or manages systems on your behalf. Ask each vendor about their own patch cadence, their security monitoring, and their incident response process. Request evidence, not assurances. If a vendor cannot show you a security roadmap or provide a recent penetration test summary, consider that a gap.
Establish a patch timeline that matches the threat. AI agents do not wait weeks to exploit a disclosed vulnerability. Set a target of applying critical patches within one week of release. For systems directly exposed to the internet, shorten that window further. If your IT team or provider cannot meet that schedule, ask what resources or process changes would make it possible.
Finally, build an AI use policy before your employees build dependencies on unapproved tools. Define what counts as sensitive data, which AI platforms are approved, and what happens when someone pastes a customer list into a public chatbot. Make the policy clear, enforce it with technical controls where possible, and train your team on it twice a year.
Why AI security is infrastructure protection, not a technology question
Horne’s warning positions AI cyberattacks on critical infrastructure as a national security issue. For your business, the same logic applies. The systems that keep your operations running, the vendors that process your payroll, the software that tracks your customer data, all of these are infrastructure. When an AI-powered attack hits one of them, your response time will be measured in hours, not days.
Build your defenses with that window in mind. Document who is responsible for applying patches, who receives security alerts from your vendors, and who has authority to shut down a compromised system. Test your incident response plan against a scenario in which an attacker moves faster than your team can convene a meeting.
The attackers will use AI to scale their reconnaissance, automate their exploits, and compress their timelines. Your preparation has to match that speed, or you will spend the days after an attack explaining to clients, regulators, and insurers why you were not ready.
How to govern AI tools your employees are already using
AI adoption in small businesses usually starts without a formal decision. Someone in sales tries a chatbot to draft emails. An analyst pastes financial data into a generative model to summarize a report. A developer uses an AI coding assistant to refactor a script. None of these actions required approval, and most of them happened before your organization had a policy.
Start by asking your team what AI tools they use, when they started, and what data they have shared. Frame the conversation as risk management, not enforcement. The goal is visibility, and you will get more honest answers if people do not fear being reprimanded.
Once you have a list, classify each tool by the sensitivity of the data it processes. A chatbot that drafts marketing copy carries different risk than one that summarizes client contracts. A coding assistant that suggests syntax changes carries different risk than one that accesses your entire codebase.
For high-risk tools, require a formal approval process. That process should include a review of the vendor’s data-handling practices, their security certifications, and their breach notification timeline. Ask whether data is encrypted in transit and at rest, whether the vendor uses your inputs to train their models, and whether you can delete your data on demand.
For lower-risk tools, establish guardrails. Prohibit pasting customer names, financial figures, or proprietary data into any public AI platform. Require employees to use company-approved accounts with logging enabled rather than personal accounts. Audit those logs quarterly to confirm the policy is being followed.
Finally, provide an approved path. If your team needs AI tools to stay productive, give them vetted options with clear use cases and security controls. A blanket prohibition drives use underground. A governed alternative keeps risk visible.
Frequently asked questions
Do small businesses really face the same AI cyberattack risks as critical infrastructure?
Your business may not operate a power plant or a water system, but you share vendors, software, and attack surfaces with organizations that do. When attackers use AI to scan for vulnerabilities, they do not filter by company size. They filter by exposure. If your firewall, VPN, or cloud platform has a flaw, an AI agent will find it at the same speed it finds one in a larger target. The risk is not identical, but the attack method is.
How quickly can AI-powered malware exploit a vulnerability once it is disclosed?
Assume that disclosure and exploitation can happen in the same day. Treat every public security advisory as an immediate action item. If a patch is available, apply it. If a workaround exists, implement it. If neither is possible, consider taking the affected system offline until a fix is ready. Speed is no longer a competitive advantage for defenders. It is the minimum requirement.
What should I do if my team is already using unapproved AI tools?
Inventory what is in use, assess the data each tool has accessed, and decide which tools meet your security standards. For approved tools, formalize their use with logging and access controls. For tools that do not pass review, migrate users to an approved alternative and revoke access to the unapproved platform. Notify your legal and compliance teams if sensitive data was processed through an unapproved service, and document the steps you took to contain the exposure.
How can I tell if my vendors are prepared for AI-powered attacks?
Ask each vendor about their patch management process, their security monitoring capabilities, and their incident response timeline. Request evidence such as recent penetration test results, SOC 2 reports, or ISO certifications. If a vendor cannot provide documentation or gives vague answers, escalate the question to their executive team. A vendor that treats security as optional is a vendor you should replace.
What AI controls should every small business have in place today?
Start with an approved-use policy that defines what data can and cannot be shared with AI tools. Implement logging for all approved AI platforms so you can audit usage. Require multi-factor authentication for any AI tool that accesses business data. Establish a patch schedule that applies critical updates within one week of release. Assign ownership of AI governance to a specific person or role, and review your controls quarterly.
Keep reading
Sources
Source: UK Cyber Chief Says the UK Will Face AI Cyberattacks on Critical Infrastructure by 2028
Researched and drafted with AI, then reviewed and approved by Ron Trotto before publishing.