Skip to main content
We have begun the founding chapter. Be among the first

Analysis

Email Never Had a Seal. Americans Lost $3 Billion Through the Gap Last Year.

Business email compromise cost Americans $3.05 billion in 2025, and AI now writes the fakes. The standard defenses check the server, not the person. A signature from a verified human, checked in the inbox, closes part of that gap.

By Sofia Martinez, Chief Executive Officer, and Dinesh Mendhe, Founder and Chairman, The AI LabNovember 13, 202611 min read

Disclosure. The AI Lab builds the Trust Identity Protocol and the free TIP browser plugin this article describes. The authors drafted it with AI assistance and are responsible for every sentence. The AI Lab registers the article on TIP as AI-Assisted.

Two envelopes closed with red and green wax seals, a stick of gold sealing wax, a brass seal stamp with a wooden handle and a lit tea candle on a white table.
A seal told a reader who closed a letter and whether anyone had opened it. Most email has never carried one. Photo by Simon A. Eugster, Wikimedia Commons, CC BY-SA 3.0, cropped.

For most of recorded history, a letter carried proof of who sent it. The writer pressed a ring or a stamp into hot wax, and the seal told the reader two things before a single word was read. It said who closed the letter, and it showed whether anyone had opened it since.

Email arrived in 1971, when Ray Tomlinson sent the first messages between two computers on ARPANET. It came without either kind of proof, and 55 years later most email still has neither. Exploiting that gap is now one of the most profitable crimes in the United States.

Americans reported $3.05 billion in losses to business email compromise in 2025, from 24,768 complaints to the FBI’s Internet Crime Complaint Center. Only investment fraud cost more. The same report counted 22,364 complaints that mentioned artificial intelligence, with $893 million lost. The tools that write a convincing email in any language are now free to use. The tools that show who wrote one are missing from almost every inbox.

The fraud keeps arriving, and it reads perfectly

The cases filed this fall show how ordinary the trick has become. On November 5, federal prosecutors in Manhattan charged three men with using AI to imitate the voices and faces of chief executives and law firm partners, and with tricking employees at companies around the world, among them two hospital systems, a bank and a pet food company, into wiring about $129 million. The charges are allegations. Other recent losses took nothing more than one ordinary email.

  • Pulaski County, Arkansas, about $1.9 million. In June an email asked the sheriff’s office to change the payment details for its jail’s medical contractor. Payments went to the new account until the real company asked about months of missed invoices on October 23. About $126,000 came back, the NWA Democrat-Gazette reported. “I don’t trust what anybody tells us now,” said Michael Hutchens, the county comptroller.
  • Liberty, Texas, about $600,000. The city really did owe a vendor that exact amount. The request to pay it came from a domain “designed to closely resemble the vendor’s legitimate email address,” the city said in a statement reported by Bluebonnet News on October 14.
  • Pittsburg, California, $913,839.81. One payment went to an account posing as a known city vendor in February. Police froze it fast and recovered $696,241, and investigators served 18 search warrants on 116 accounts. The city said the main suspect, based in Nigeria, had compromised its email system (city release, August 4).
  • Iowa City and Ohio, more than $2.4 million in two wires. Two men who were members of the U.S. Air Force at the time were sentenced on September 25 to a combined 189 months. They stole email passwords, then used “spoofed” addresses that mimicked victims and their business partners to redirect a wire of more than $1.68 million from Iowa City and one of more than $720,000 from Ohio.
  • Delaware, November 12. The state’s environmental agency warned permit holders about emails from addresses such as [email protected] demanding fees through a link. Some quoted real permit application numbers and staff names, likely taken from public notices.

Some of these began with a stolen password and some with a lookalike address. Every one ended the same way, with a person who believed that an email came from someone they trusted.

Your inbox checks the server, not the person

Three standards guard most email today. SPF lists the servers allowed to send mail for a domain. DKIM lets a domain sign the mail it sends. DMARC tells a receiving server what to do when those checks fail. Since 2024 Google and Yahoo have required bulk senders to use them, and Microsoft added similar rules for high-volume senders to Outlook.com in May 2025. They stop a great deal of crude forgery.

All three answer one question, which is whether a server was allowed to send mail for a domain. None of them says which human wrote the message. Three gaps follow from that.

  • The lookalike domain. A criminal who registers a name one letter away from a supplier’s can set up SPF, DKIM and DMARC for it, and the mail passes. The DMARC standard, RFC 7489, says it does not address “visually similar domain names.”
  • The display name. Most mail apps show a name and hide the address behind it, especially on a phone. The same standard says it does not address abuse of that name either.
  • The hacked account. When a criminal logs in to a real mailbox, every message sent from it is authentic by the standards’ own definition, because it comes from the right server.

Spelling mistakes used to give fakes away. Chat tools have removed that clue, which leaves the reader with a name, an address and a tone of voice, all of which can be copied.

A signature from a person, checked where the email is read

The AI Lab built email signing into the free TIP browser plugin, part of the open Trust Identity Protocol. It works in Gmail, Outlook, Yahoo Mail, AOL Mail, iCloud Mail, Proton Mail, Zoho Mail and Fastmail, in Chrome, Brave, Arc and Firefox on a computer.

The sender writes an email as usual and presses Sign with TIP, next to Send. A fresh fingerprint or face check on the sender’s own device signs that one email, and two short lines go under the message, “Signed with TIP®” and a Verify link. The signature covers the subject, the words, the sender’s address and the people it was sent to. A second email needs a second check. Nothing stays unlocked in between.

How a signed email looks to the sender, to the reader, and to the reader of a fake. The wording matches what the TIP plugin shows today.

A reader who has the TIP plugin sees the result the moment the email opens, above the message, with nothing to click. A genuine email reads “Signed by” and the sender’s verified name. If anyone changed the words, or pasted the TIP line under a different message, it reads “Changed since” that person signed it, with the date.

A reader without the plugin opens the Verify link instead. It leads to a public record showing who signed the email and when, and the reader can paste the email there to check that it matches, free and without an account. On a phone the link works the same way.

The record proves the email without holding it

A public record that kept people’s emails would itself become a target, so TIP keeps none. When an email is signed, its subject and words travel over an encrypted connection to a TIP node, which uses them only to compute a fingerprint of the text and keeps no copy. The public record holds that fingerprint, sealed commitments to the sender’s address and to the recipients, the sender’s TIP-ID and the time. It never holds the subject, the words or anyone’s address, and no address leaves the sender’s browser. Bcc recipients are sealed with a secret that stays on the sender’s device, so the other recipients cannot tell who was copied. The TIP Privacy Policy, section 3.6, sets this out in full.

The email itself is the evidence. Whoever holds it can check it against the record, and nobody without it can learn what it said.

The missing signature becomes the warning

One signed email proves little on its own. The value appears when signing becomes a habit for the emails that move money.

If a title company signs every set of wiring instructions, or a finance team signs every change of bank details, a buyer who receives unsigned instructions in that company’s name has a reason to pick up the phone. A criminal sitting inside the company’s hacked mailbox can send from the right address, but cannot produce the signature, because the signing key stays encrypted on the employee’s own device and opens for one signature at a time.

The web went through the same shift. Once most sites used encrypted connections, Google’s Chrome began labeling every unencrypted page “Not secure” in July 2018. The padlock stopped being a feature people noticed and its absence became the thing that stood out. Email can follow for the messages that move money.

A criminal inside a hacked mailbox can send from the right address. The signature is the one thing they cannot send.

What a signature will not do

  • It does not block a fake from arriving. Keep the spam filter, SPF, DKIM and DMARC. TIP adds a check a person can see. It filters nothing.
  • It proves who signed and that nothing changed, not that the email is true. A signed request for a wire still deserves the usual call to a known number.
  • It only helps readers who look. Without the plugin, a reader has to open the link, and some never will.
  • Email signatures have failed before. S/MIME and PGP have existed since the 1990s and never reached most inboxes. In a 1999 study, “Why Johnny Can’t Encrypt”, most participants given 90 minutes could not sign and encrypt a message with PGP 5.0. TIP removes the steps that defeated them. Identity is verified once, there are no certificates to buy or swap, signing is one press and a fingerprint, and checking is a link anyone can open. Whether people adopt it is the open question, and only use will answer it.
  • It runs in webmail on a computer today. Phone mail apps open the Verify link but do not show the in-inbox check. The Safari version with email signing is with Apple for review. Chrome and Firefox have it now.

A checklist for CEOs, finance teams and anyone who pays invoices by email

Most of this works without TIP. Forward it to whoever pays the bills where you work.

  1. Confirm any change of bank details by phone. Use a number you already had, from a contract, an old invoice or the company's website. Never use a number or link from the email that asks for the change.
  2. Read the address, not the name. Tap or hover on the sender's name to see the full address. Look for one swapped letter, an extra word, or a different ending such as .co instead of .com.
  3. Treat urgency plus secrecy as a stop sign. "Today", "before the auditors see it" and "keep this between us" are how these emails push people past their own rules.
  4. Make the callback rule written policy, signed by the CEO. Staff should never have to choose between an urgent email from the chief executive and a phone call to check it. A written rule that every payment change gets a callback, backed by the CEO, removes that choice.
  5. Require two people for a new payee. Any new account or changed account gets a second approver who did not receive the request.
  6. Set DMARC to reject for your own domain. A DMARC policy of p=reject stops criminals from sending mail that uses your exact domain. Ask whoever runs your email to check it.
  7. If money went out, act within hours. Call your bank and ask it to recall the transfer, then file at ic3.gov. In 2025 the FBI froze $679 million for victims who reported quickly, 58% of the money at risk in the cases it pursued.
  8. Sign the emails that move money. With TIP, the people who pay you can check that an invoice or a change of details came from you, unchanged.

How to sign your first email

  1. Get a free TIP-ID at vp.theailab.org. It takes a one-time identity check of about a minute.
  2. Add the free TIP plugin from the Chrome Web Store (Chrome, Brave and Arc) or Firefox Add-ons.
  3. In the plugin’s Settings, open Activate email signing and press Turn on all. Your browser asks once.
  4. Write an email in your webmail and press Sign with TIP.

Checking a signed email is free for everyone. Signing is free for individuals, nonprofits, schools, governments, journalism organizations and businesses under USD 100,000 in annual revenue, and from December 14, 2026 Free includes 100 signed emails a month, with 500 on Plus, 1,000 on Pro and 5,000 on Max. Larger businesses use the published commercial license, with every tier on the page. To plan a rollout for a finance team or a whole company, write to [email protected].

For the full guide to spoofed email, read how to protect yourself from email spoofing, and for payments teams, how finance teams stop business email compromise. When an outage or a breach sets off a wave of fake “help” emails, read how to know what’s true in a crisis.

Frequently Asked Questions

How can I prove an email came from me?

Sign it with a key only you control. With the free TIP browser plugin and a TIP-ID, you press Sign with TIP next to Send and a fresh fingerprint or face check signs that one email. Anyone who received it can check your verified name and that the email has not changed.

Can someone copy the TIP line onto a fake email?

They can copy the line, but not the signature behind it. The check compares the email in front of the reader with what the sender signed. A different message, or the same line under new words, reads "Changed since" the sender signed it.

Does TIP read or store my emails?

No copy is kept. The subject and message go over an encrypted connection to a TIP node, which uses them only to compute a fingerprint. The public record holds that fingerprint and sealed commitments, never the words, the subject or anyone's address.

Do the people I email need anything installed?

No. Anyone can open the Verify link under the message, see who signed it and paste the email to check it, free and without an account. People who have the TIP plugin see the result in their inbox without clicking.

Which email services and browsers does it work with?

Gmail, Outlook, Yahoo Mail, AOL Mail, iCloud Mail, Proton Mail, Zoho Mail and Fastmail, in webmail on a computer, with Chrome, Brave, Arc or Firefox. The Safari version with email signing is with Apple for review.

Is DMARC enough to stop business email compromise?

No. DMARC at p=reject stops forged mail that uses your exact domain, and every company should set it. Its own standard says it does not address lookalike domains or display names, and mail sent from a hacked real account passes it.

How can a company protect itself from CEO fraud emails?

Make a phone callback on a known number mandatory for every payment change, require two approvers for new payees, set DMARC to p=reject, and have the people who move money sign their emails, so an unsigned payment request in their name stands out.

Does this replace SPF, DKIM and DMARC?

No. Those standards check that a server may send mail for a domain, and they block a lot of forgery. TIP adds a check of the person who signed an email and whether it changed. Keep both.

What does email signing cost?

Checking a signed email is free for everyone, and the plugin is free from each browser's official store. Signing is free for individuals, nonprofits, schools, governments, journalism organizations and businesses under USD 100,000 in annual revenue. From December 14, 2026, Free includes 100 signed emails a month, Plus 500, Pro 1,000 and Max 5,000. Larger businesses use the published commercial license at theailab.org/tip-license, which includes 5,000 a month.

Sources


Sofia Martinez is the Chief Executive Officer of The AI Lab Intelligence Unobscured, Inc., which created the Trust Identity Protocol, and leads the standards, ethical frameworks and product execution behind TIP. Dinesh Mendhe is its Founder and Chairman. Hero photograph by Simon A. Eugster, licensed under CC BY-SA 3.0, cropped; the crop is shared under the same license.