AI Phishing: How Deepfake and Voice Cloning Scams Target Businesses

November 20th, 2026
AI Phishing: How Deepfake and Voice Cloning Scams Target Businesses

Phishing used to be easy to spot. Bad grammar, a strange greeting, a "prince" who needed help moving money. Most employees learned to delete those emails without a second thought.

AI phishing changed that. Attackers now use the same AI tools businesses rely on to write clean, convincing emails in seconds. They clone voices from a short audio clip. Some even fake live video calls. The FBI's 2025 Internet Crime Report tracked AI-related fraud for the first time and found $893 million in losses tied to AI, a figure the FBI said is probably understated.

That's a big number. But the bigger problem is that most businesses still train employees to spot the old kind of scam. Would your team recognize the new one?

What Is AI Phishing?

AI phishing is any phishing attack that uses artificial intelligence to trick someone into sharing information, clicking a link or sending money. The goal hasn't changed. Attackers still want passwords, payments and access. What changed is how good the bait looks.

It comes in a few forms. AI-written emails and texts. Cloned voices on the phone. Deepfake video on a call. Often an attack mixes them, starting with an email and following up with a "quick call" from someone who sounds exactly like the boss.

Phishing was already the front door for most attacks, long before AI showed up. AGJ wrote about a phishing attack at a Gulf Coast hospital that put about 30,000 patients' data at risk through one employee's email account. AI phishing makes that kind of attack easier to pull off and harder to catch.

How AI-Generated Phishing Emails Fool Employees

AI-generated phishing emails read like real business email. The grammar is clean. The tone matches the sender. Attackers can feed an AI tool a company's website, a LinkedIn profile and a few public posts, then ask it to write a message that sounds like the CFO.

Business email compromise was already one of the most expensive crimes the FBI tracks. Its 2025 report put business email compromise losses at about $3 billion, out of nearly $21 billion in total reported cybercrime losses. AI makes these attacks faster to write and easier to tailor. An attacker who once sent the same clumsy email to 1,000 people can now send 1,000 personal ones.

The classic CEO scam email still works the same way. An urgent request from leadership, a vendor payment that needs to go out today, a gift card purchase for a client. The difference is that nothing about the message looks off anymore.

Voice Cloning Scams and Deepfake Calls

Email is only part of it. AI phishing has moved to the phone and to video, and honestly, this part worries security teams the most.

Voice Cloning Scams

Voice cloning scams use a few seconds of recorded audio to copy someone's voice. A voicemail greeting, a conference video or a podcast clip is usually enough. The FBI has warned about an ongoing campaign that uses AI-generated voice messages to impersonate senior U.S. officials, and the same technique works against any business.

Picture a bookkeeper getting a call that sounds exactly like the owner. He's traveling, the connection is a little rough, and a vendor needs a wire before 3 p.m. Most people would want to help. That's what the attacker counts on.

Deepfake Video Calls

Deepfake scams take it a step further. In early 2024, an employee at engineering firm Arup joined a video call with what looked like the company's CFO and several coworkers. Every one of them was fake. The employee sent about $25 million across 15 transactions before checking with the head office.

A small business probably won't lose $25 million. But a $40,000 wire can hurt a 20-person company far more. And small offices tend to have fewer approval steps, which can make them easier targets.

Warning Signs of AI Phishing

Spelling mistakes aren't a reliable clue anymore. The better signs of AI phishing have more to do with behavior than with how a message looks:

  • Urgency, especially around payments, passwords or gift cards
  • A request to keep something quiet or skip the normal process
  • A sudden switch to a new phone number, personal email or messaging app
  • Small differences in an email address, domain or phone number
  • On calls, odd pauses, flat emotion or a voice that sounds slightly off
  • On video, stiff movement, blurry edges or lighting that doesn't match

The FBI's advice is simple. Don't assume a message is real just because it sounds right. Verify it through a separate channel.

How to Protect Your Business From AI Phishing

No single tool stops AI phishing. A few layers, used together, make a big difference.

  1. Set a callback rule. Any request to move money or change payment details gets verified by calling a known number, not the one in the message. Put it in writing so employees feel safe pausing a request from the boss.
  2. Require two approvals for payments. A second person on every wire transfer stops most of these scams cold.
  3. Use a code word. Leadership and finance staff can agree on a phrase to confirm identity on unexpected calls. It feels a bit silly at first. It works.
  4. Turn on phishing-resistant MFA. CISA recommends phishing-resistant multifactor authentication because it protects accounts even when someone hands over a password.
  5. Update security training. Show employees real examples of AI phishing, including cloned voices. Old training built around typos probably won't help much.
  6. Filter email at the gateway. Modern filtering catches spoofed domains and suspicious links before they reach an inbox.

It's worth limiting what attackers can learn, too. Long voicemail greetings and public org charts give them material to work with. That's a smaller step, but perhaps an easy one.

Planning for the Attack That Gets Through

Even good defenses miss sometimes. One click on the wrong link can let ransomware in, and AI phishing makes that click more likely. That's why recovery planning belongs in the same conversation.

Tested backups, a written incident response plan and a clear list of who to call can turn a crisis into a bad week. AGJ covers the basics of building a data recovery plan before disaster strikes, and it's a good place to start.

AI phishing is one piece of a larger set of risks tied to AI in the workplace. Businesses that address it now, before an incident, usually spend far less fixing it later.

Getting Help With AI Phishing Protection

Most small businesses don't have a security team to set up email filtering, MFA and staff training on their own. AGJ Systems helps businesses across the Gulf Coast put those protections in place through its IT and network security services.

A free 15-minute discovery consultation is a simple way to find out how well the business would hold up against AI phishing.

Frequently Asked Questions

What is AI phishing?

AI phishing is a phishing attack that uses artificial intelligence to trick people into sharing information, clicking links or sending money. Common forms include AI-written emails, cloned voices on phone calls and deepfake video calls.

How do voice cloning scams work?

Voice cloning scams use a short audio sample, such as a voicemail greeting or online video, to copy a person's voice with AI. Attackers then call employees or family members while posing as that person, usually to request money or sensitive information.

How can you spot a deepfake scam?

Watch for urgent requests, pressure to skip normal procedures and sudden changes in contact details. On video or audio, look for stiff movement, odd pauses, flat emotion or mismatched lighting. Always verify unusual requests through a separate, known channel.

Can spam filters stop AI-generated phishing emails?

Email filtering catches many AI-generated phishing emails, especially those with spoofed domains or malicious links. Some still get through, so businesses should pair filtering with phishing-resistant multifactor authentication, callback rules for payments and regular employee training.

What should a business do after an AI phishing attack?

Contact the bank immediately if money was sent, reset affected passwords, notify the IT provider and report the incident to the FBI's Internet Crime Complaint Center at ic3.gov. A written incident response plan and tested backups make recovery much faster.