For more than a decade, the messaging industry has fought SMS fraud by guarding the point of entry. We verify the sender, check the tax ID, confirm the business address, and add the brand to a registry. I have spent years working alongside the people who build those registries, and I respect what they have achieved. I also believe the industry has spent all its energy on one door while leaving the rest of the house unlocked.
AI has made that gap impossible to ignore. Fraudsters no longer need a person at a keyboard to hijack a message or impersonate a bank. A bot can do both at a scale no human team could match, and the traditional verification process was never designed to catch it.
Sender ID Registries Solved the Problem We Had Ten Years Ago
When smishing took off more than a decade ago, most attacks followed the same pattern. A criminal posed as a bank or an online retailer, and later as a crypto wallet provider, and asked the recipient for login details or card numbers. The logical response was to make sure only the real company could send messages under its own name.
Sender ID registries grew out of that logic. About 60 countries now require businesses to register their sender IDs before messaging consumers, and the list keeps growing. Registration usually means checking tax IDs, confirming the business address, and verifying that the company actually exists.
The model works best when the senders take part themselves. The UK registry is the strongest example. The Mobile Ecosystem Forum built it together with regulators, networks, major UK financial institutions. Other banks joined, along with roughly 100 to 150 other large companies. Those organizations signed up because they understood the threat they were facing.
Even a strong registry only verifies one moment in the life of a message. Faking a business online has never been hard. I have watched fraudsters create dozens of fake local business listings on Google Maps, each one pinned to an empty lot. A registry checks the sender at the moment of entry, and fraud can happen at any point after that.
Two Attacks That AI Has Made Cheap and Fast
Man in the middle
The first attack targets the message while it travels in the telecom network. While a WhatsApp message stays encrypted from the moment it leaves the sender until the WhatsApp app on your phone decrypts it, an SMS message has no such protection. An SMS crosses the IP network in the open, and the players on that network are largely anonymous. Anyone with moderate software skills can access te network.
Once inside, a fraudster can hijack traffic in transit. Picture a bank sending a one-time password to a customer. An attacker who can identify that bank’s OTP traffic can intercept the message, replace the legitimate link with one that leads to a spoofed login page, and pass it along. The customer sees a familiar sender and a familiar format, and has no reason to suspect anything. We see the same trick with fake delivery notifications copied from real courier messages.
For years, the industry’s expectation was that we would be able to use AI to spot anomalies in the traffic despite the inherent weakness of that approach. Detection starts 100 meters behind the starting line, because by the time a pattern looks anomalous, a great deal of fraudulent traffic has already gone through. Now the attackers have AI as well. Bots can identify targets, and rewrite or reroute messages automatically on a massive scale, with no human pressing a button. Attacks of this kind have rapidly escalated over the past year and a half, and the industry has no answer for them yet.
Identity spoofing
The second attack is the oldest trick in the book, rebuilt for a new era. A criminal claims to be a bank or a major brand, and AI now makes that fake identity convincing enough to slip past traditional verification. My panel colleagues lead companies at the center of this work. Eli manages a global number verification provider, and John handles carrier verification for North American networks, including 10DLC registration. They acknowledge that AI is spoofing its way into sender ID registries, and their focus is on closing that hole.
I agree the hole needs closing. But simply closing one gaping hole leaves the transit problem unsolved, because right now nobody is dealing with it.
Spoofing happens on the receiving end too. AI fraudsters invent fake customers, compromise or clone a SIM card, or fake the IMEI number that identifies a handset. Verifying who receives a message matters as much as verifying who sends it.
We need a holistic approach.
Who Pays When SMS Fraud Works?
Consumers pay first, and the economics explain why these attacks never stop. Take Israel, a country of about 9 million people, and imagine a bank that holds 30% of the market. A fraudster who sends a million messages claiming to be that bank will reach around 300,000 of its real customers by simple odds. If only 5% of those customers are naive enough to respond, the fraudster walks away with access to 15,000 bank accounts. Sending a million messages costs very little. As long as that math holds, the business model will survive.
Businesses pay second, and the damage spreads to the whole industry. AI fraudsters create fake users, use bots to trigger OTP messages in a company’s name, and route the traffic to a terminating partner who collects the fees. After Elon Musk bought Twitter, he said the company had been losing tens of millions of dollars a year on verification messages generated this way. Those messages were never intended for a real person.
A company that discovers it has been paying for fraudulent traffic often stops trusting SMS altogether. The hyperscalers have already moved on. Meta verifies accounts through its own apps, and Google and Apple rely on passkeys and their own account channels. These operational changes were a reaction to SMS fraud, and every one took volume away from the channel.
Why the Message Itself Needs Protection
Over the last 5 years our position has been simple and consistent. It’s become even more urgent as AI tools have become accessible. Verifying the sender is necessary, but verification alone will never be enough. As long as SMS messages transit through the open network, bots can lift them off the wire and repurpose them. The only real defense against that kind of attack is to encrypt messages at the source, rendering them unreadable and unaccessible.
RCS shows both the promise and the limits of a newer channel. RCS runs through the native messaging app on most phones, and as it is encrypted through most of its journey, has stronger protections than SMS. RCS business messaging also has a sender verification process of its own, but a determined fraudster can still get around it. With AI, anyone can design a message that looks exactly like it came from a bank, send it from a personal account instead of a verified business account, and reach a million people without passing any verification at all.
Switching channels does not remove the need to protect the whole journey, from the sender to the screen.
What Messaging Providers Owe Their Customers
Aggregators like MMDSmart do not create these threats, but we carry real responsibility for protecting our customers and the people they message. That responsibility starts with transparency. You deserve to know what threats exist, how they work, and what alternatives are open to you.
Responsibility also means building our own defenses rather than waiting for the industry to agree on a standard. MMDSmart runs its own filtering and detection systems, designed to identify traffic that fraudsters generate artificially and to catch cases where someone is spoofing a customer’s account. Not long ago, our team noticed messages flowing under one customer’s name that the customer had never sent and flagged them as anomalous behavior, saving our client a significant amount that would have been charged to them if we had not interceded on their behalf.
Finally, responsibility means honest advice about channels. When you ask how to reach your audience, a good provider should tell you which channel fits the message, which one carries the lowest risk, and how to get real value for what you spend. Sometimes the right answer is SMS. Sometimes the right answer is an interactive WhatsApp message or a verified RCS message.
I will be making this case on a panel on The MEF Connects Anti-Fraud Forum October 28, alongside industry leaders with whom we have worked on this problem for years. My challenge to the industry is direct. We have spent a decade protecting the front door, but AI fraud is already moving through it and every other part of the message journey. We need a holistic approach. Refitting sender registration against AI fraud is overdue, and so are protecting the message in transit and ensuring consumer integrity.
If you want to understand how exposed your own messaging traffic is, talk to the team at MMDSmart MessageWhiz.