The voicemail sounds like your owner and a request you’ve heard a dozen times: wire the deposit to the new vendor account before the bank closes. Nothing about the message seems wrong, except that your owner never left it.
AI has made that kind of deepfake cheap to produce, along with error-free phishing emails and automated scans for unpatched software.
Eight controls protect your business from AI cyberattacks, and the first three target the payment fraud and account takeovers behind some of the largest business losses.
Key Takeaways
- AI lets attackers run familiar scams and break-ins at larger scale and with fewer of the mistakes that used to give them away, which shortens the time businesses have to put known defenses in place.
- A familiar voice, a video call, or a well-written email no longer proves identity. Any request to move money or change account access needs a callback to a number already on file.
- Attackers call help desks and IT providers posing as employees to get passwords and MFA reset, so identity checks have to cover your own staff too.
- Many software flaws are now exploited before a patch exists, which makes patching speed, internet-facing exposure, and 24/7 monitoring the deciding factors.
- AI agents can carry out much of an attack with little supervision, but every documented case so far was aimed and directed by people.
What Counts as an AI-Powered Cyberattack
An AI-powered cyberattack is any attack in which the criminal uses artificial intelligence to do part of the work. That can mean writing the phishing message, cloning a voice, finding a weak spot in your software, or carrying out steps of the break-in.
The goal has not changed with AI: attackers still want money, data they can sell, or access they can hold for ransom.
Verizon’s 2026 Data Breach Investigations Report found that, in the median case, attackers used generative AI for 15 distinct attack techniques. Verizon’s analysts concluded that AI’s main impact so far is “automating and scaling techniques defenders already know how to detect.”
The techniques are familiar, so the defenses against them already exist. Attacks now arrive faster and in larger numbers, which means slow or partial defenses fail more often.
What Makes a Business an Easy Target
Automation lets attackers test thousands of businesses at once, so your company does not need to be large or well known to be picked.
AI-assisted campaigns go after whatever weakness they find first. The businesses that get hurt usually have one or more of these gaps:
- Email and banking logins protected by a password alone, or by text-message codes
- One person who can send a wire or change vendor bank details without a second approval
- Firewalls, VPNs, or websites that go weeks without security updates
- No one watching for suspicious activity at night or on weekends
- Vendors and former employees with access nobody has reviewed
- Backups that have never been tested with a full restore
Your exposure also runs through the companies you work with. Verizon found that 48% of breaches in its 2026 report involved a third party, up 60% from the year before.
For a businesses, that third party is usually a payroll platform, an accounting firm, a software vendor with remote access, or the IT provider.
How Attackers Use AI Against Your Business
Outside attackers use AI against businesses in five main ways, all of them versions of attacks that existed before AI.
Phishing That Reads Like a Coworker Wrote It
The phishing email your staff learned to spot, with its odd grammar and generic greeting, is disappearing. The FBI has warned that criminals use AI to write messages faster, reach more victims, and remove the spelling and grammar errors that used to give scams away.
Google’s Threat Intelligence Group found that attackers also use AI for research, pulling together public information to identify decision-makers and map how a target organization is structured.
Texts and messaging apps carry the same risk with fewer safeguards. Verizon found that social engineering delivered through mobile channels had a 40% higher median click rate than attempts that arrived by email.
Voices and Faces Your Staff Already Trust
Voice cloning and real-time deepfake video let an attacker sound and look like your owner, your CFO, or a longtime vendor. The FBI lists cloned voices and deepfake video calls impersonating executives among the AI tactics criminals now use against victims.
The clearest documented case involves the engineering firm Arup. In January 2024, a finance employee in its Hong Kong office joined a video call with what appeared to be the CFO and several colleagues. Every other person on the call was an AI-generated fake, and the employee sent 15 transfers totaling about $25.6 million.
Smaller businesses rarely make their losses public, but the FBI’s totals show the scale. Business email compromise, the category these payment schemes fall under, produced more than $3 billion in reported losses in 2025, according to the FBI’s 2025 Internet Crime Report.
Calls to Your Help Desk Posing as Your Employees
Some attackers skip your staff and go after the people who reset your staff’s passwords.
Mandiant’s M-Trends 2026 research found voice phishing was the most common way into cloud environments, behind 23% of the cloud intrusions Mandiant investigated.
CISA described the pattern in its advisory on the Scattered Spider group. Callers pose as employees and talk IT or help desk staff into resetting passwords and moving MFA to a device the attacker controls. They gather names, roles, and personal details from social media and company websites first.
Voice cloning makes that impersonation easier, since a caller can now sound like the employee they claim to be. If your business uses an outside IT provider, ask how it confirms a caller’s identity before a reset. The answer should involve more than a name and a birth date.
Software Flaws Exploited Before You Can Patch
AI shortens the time between a software flaw becoming known and criminals exploiting it, with a mean exploitation time of seven days.
Most businesses without managed cybersecurity are not patching at anything close to that speed.
Verizon’s data shows exploiting vulnerabilities grew to 31% of initial access in breaches, a 55% jump from the prior year, while the median time to fully fix a critical vulnerability rose to 43 days.
Automated Attacks Run by AI Agents
An AI agent is software that takes a goal, breaks it into steps, and uses tools to carry out those steps with little supervision. Attackers now use agents to do the labor of an intrusion.
In a case Google published in September 2026, a financially motivated attacker breached an organization’s cloud environment and deployed an autonomous agent framework. The attacker then planned, built, and ran a mass credential-harvesting campaign in under six hours.
In a 2025 campaign that Anthropic disrupted, AI performed an estimated 80 to 90% of the work against about 30 organizations.
Neither attack involved a rogue AI choosing its own victims; people picked the targets and set the goals. The agents removed the staffing limit that used to cap how many organizations one group could work against at once, and they keep working overnight and on weekends.
How to Protect Your Business From AI Cyberattacks
The eight controls below are listed in priority order. The first three target payment fraud and account takeover, the next two close the speed gap, and the last three protect your recovery and your company’s name.
Our post on the layered approach to cybersecurity explains why no single control is enough on its own.
1. Set a Verification Rule for Money and Account Changes
Confirm every request to send money, change bank details, reset a password, or grant access by contacting the person yourself, using contact details you already have.
Hang up and call back on a number already on file, or message the person on a separate platform.
Add a second approver for wires and vendor banking changes above a set amount. If one person in your office can send a wire on their own authority, fix that first.
Apply the same rule to password and MFA resets, including any your outside IT provider handles. Write the rule down and build it into your security awareness training for employees.
Make the fraudster’s research harder by reviewing what your website and staff profiles reveal about who approves payments, which vendors and banks you use, and how to reach leadership directly.
2. Move to Phishing-Resistant Multi-Factor Authentication
Text message codes and simple push approvals can be phished or worn down by repeated prompts. The Cybersecurity and Infrastructure Security Agency (CISA) calls phishing-resistant multi-factor authentication (MFA), based on the FIDO/WebAuthn standard behind passkeys and hardware security keys, the “gold standard” for protecting accounts.
CISA recommends rolling it out in phases, starting with high-value accounts. For most businesses, that means email administrators, anyone who approves payments, and executives.
Where a system does not support phishing-resistant MFA yet, CISA recommends number matching on push prompts as an interim step.
3. Limit Who Has Access, Including Vendors
When an attacker or an AI agent steals one password, the damage depends on what that account can reach. Give each employee access to the systems their job requires and nothing more. Keep administrator rights to a small number of separate accounts that are not used for everyday email and browsing.
Review vendor access the same way. Ask every vendor with access to your data or systems whether it enforces MFA, how fast it applies security patches, and how quickly it will report an incident to you. Remove accounts and remote access tools for vendors and former employees you no longer work with.
4. Patch Faster and Shrink What Faces the Internet
Turn on automatic updates for every laptop, phone, and web browser. Apply critical patches to firewalls, VPNs, and remote access tools within days, since internet-facing systems are the easiest for automated scanning to find.
Australia’s national cyber agency singled out websites in July 2026, warning that attackers may use AI to rapidly find weaknesses in them. Its small business guidance suggests asking whoever maintains your site these questions:
- Is the software behind our website still supported with security updates?
- Are updates applied automatically, or quickly after release?
- Is the software securely configured, with unneeded features turned off?
- If the site is hacked, how fast would you notice, tell us, and restore it from a clean backup?
5. Put Someone on Watch Around the Clock
The median handoff from the criminal who breaks in to the group that exploits the access took 22 seconds in 2025. Paired with agent-run campaigns that finish in hours, those numbers mean a weekly log review finds the damage after it is done.
Effective coverage means 24/7 proactive monitoring with someone authorized to isolate a device or disable an account at 2 a.m.
6. Keep Backups an Attacker Cannot Reach
Ransomware remains the payoff for many AI-assisted intrusions, and attackers may target backups first so you cannot recover without paying. Keep at least one copy offline or immutable, meaning it cannot be changed or deleted for a set period.
Test a full restore on a schedule, since a backup job that reports success can still fail when you need the data back. Our data backup and disaster recovery service is built around that test.
7. Lock Down Your Email Domain
Attackers also impersonate your business to your customers and vendors. Three email authentication records, SPF, DKIM, and DMARC, let receiving mail servers confirm that a message claiming to come from your domain was sent by you.
The FTC recommends confirming that your email provider supports all three and can set them up correctly, since a misconfiguration can block legitimate mail. A DMARC policy set to reject spoofed messages makes it much harder for a fraudster to send a fake invoice under your company’s name.
8. Write Down What Happens When Something Goes Wrong
Only 34% of small and midsize businesses have a formal incident response plan, according to a Guardz survey cited by the U.S. Chamber of Commerce. Without one, the first hours of an attack go to figuring out who to call.
A workable plan can fit on a few pages, naming who makes decisions, how to isolate affected systems, who calls the bank if money moved, how to restore from backup, and what you tell employees and customers.
Do You Need AI-Powered Security Tools?
AI-driven security tools help most in monitoring, where they can sort large volumes of alerts and flag unusual behavior faster than people can.
Michael Siegel, director of Cybersecurity at MIT Sloan, has argued that “AI-powered cybersecurity tools alone will not suffice.” In his view, defense needs human oversight and governance alongside automation.
A product labeled “AI” will not help if your email still runs on passwords alone or your firewall is months behind on patches. When a vendor pitches one, ask three questions:
- What does it detect or block that our current tools miss?
- Who responds when it raises an alert at night or on a weekend?
- What does it need from us, such as MFA, patching, or logging, to work as advertised?
Where to Start Based on the Size of Your Business
The eight controls apply at every size, but the order and the help you need depend on your headcount and IT coverage. Texas SB 2610 uses similar size tiers, so the same work can count toward safe harbor protection.
Fewer than 20 employees, no IT staff
Start with the verification rule, MFA on email and banking, automatic updates, and a tested backup. SB 2610 expects documented policies, password management, security training, and reliable backup at this size.
20 to 99 employees
Add access reviews, a patching schedule with deadlines, 24/7 monitoring, and a written incident response plan. SB 2610 ties this tier to CIS Controls Implementation Group 1 (IG1), a set of 56 foundational safeguards from the Center for Internet Security.
100 or more employees
Align your program with a recognized framework such as the NIST Cybersecurity Framework or the full CIS Controls, and require identity verification for every help desk password and MFA reset. Keep administrator accounts separate from everyday accounts, and review vendor access at least once a year.
An internal IT person or team
The usual gap is overnight coverage and security specialization. Co-managed IT services can add overnight monitoring and incident response without replacing your staff.
Our managed cybersecurity services cover monitoring, patching, MFA rollout, and the documentation behind them for businesses across Lubbock, Amarillo, Midland, and Plano. If you are not sure where your gaps are, a free IT assessment is a practical place to start.
Conclusion: Business Protection From AI Cyberattacks Now Has to Match the Speed of the Attack
AI has removed much of the time and skill that limited how many businesses an attacker could reach, and how effectively.
Cybersecurity defenses that once had weeks to catch up now have days or hours, and a familiar voice or polished email can no longer be trusted to confirm who is asking.
This week, put your callback rule for payments, bank changes, and account resets in writing. The rule requires little beyond a documented policy and staff training, and it addresses business email compromise, which produced more than $3 billion in reported losses in 2025, according to the FBI.
FAQ
Is My Business Safe From AI Cyberattacks?
No business is safe by default, but most AI-assisted attacks still rely on familiar weaknesses: stolen passwords, unpatched software, and staff who act on a convincing request. Phishing-resistant MFA, fast patching, 24/7 monitoring, and a strict rule for verifying payments make your business a much harder target.
Can a Rogue AI Agent Attack My Business on Its Own?
No documented case so far has involved an AI agent attacking on its own. The reported AI-agent attacks, including campaigns described by Anthropic and Google, were set up and directed by people who chose the targets. The agents handled much of the work, which lets a small group go after many more businesses at once.
Will Antivirus Stop AI-Powered Attacks?
Antivirus alone will not stop them. Google has documented attackers using AI to produce malware that keeps changing its code to avoid detection, which weakens tools that look for known signatures. Endpoint detection and response software, which watches for suspicious behavior, paired with people who act on its alerts, does far more.
How Fast Do You Need to Install Security Updates?
Critical patches for internet-facing systems such as firewalls, VPNs, and websites should go in within days. Attackers now often exploit flaws before or shortly after a fix is released, so a monthly patch cycle leaves those systems exposed for weeks.
Where Should You Report an AI-Enabled Scam?
File a report with the FBI’s Internet Crime Complaint Center, and call your bank right away if money moved. A fast report gives the bank the best chance of stopping or recalling a transfer.