Published: September 27, 2026 · Last updated: September 27, 2026
TL;DR: AI voice agents are now calling businesses to check balances, file claims, and negotiate on a customer’s behalf, and most contact centers have no way to tell a legitimate agent from a malicious bot. Pindrop’s new BotStopper technology identifies whether a caller is human or machine in about two seconds. One Fortune 500 healthcare company using it cut bot call volume by 94.3% in four months, after AI-powered attacks against contact centers rose nearly 1,680% since late 2024.
Picture a customer service line at a bank. Some of the calls coming in are not people anymore. They are AI agents, dialing in on behalf of a customer, or pretending to.
That is not a hypothetical. It is happening at scale right now, and most contact centers have no reliable way to tell the difference.
The Problem Nobody Built a Front Door For
Voice security company Pindrop has spent years building tools to catch deepfake audio and synthetic voices in call centers. Its newest product, BotStopper, tackles a related but distinct problem: not whether a voice sounds fake, but whether the thing on the other end of the line is an AI agent at all, and if so, which one, working for whom.
The scale of the shift is what makes this urgent. Pindrop found AI-powered attacks against contact centers increased nearly 1,680% since late 2024, across more than 700 million customer interactions it analyzed. Research firm IDC projects 1.3 billion AI agents will be operating by 2028. Some of those agents will be doing legitimate work, booking appointments or checking account balances for the person who owns them. Others will be probing systems for information, tying up staff time, or impersonating real customers outright.
Two Seconds to Decide
BotStopper is built on Pindrop’s existing Pulse deepfake detection engine, and it works fast: the company says it can identify whether a caller is human or machine within roughly two seconds of the call connecting. That speed matters, because a contact center cannot put a customer on hold for a minute while it runs a background check on who is calling.
“The goal is to know which agent is calling, who it represents and what it should be allowed to do,” said Vijay Balasubramaniyan, Pindrop’s CEO.
That sentence is doing more work than it looks like. Detecting a bot is step one. Knowing which specific agent it is, and what permissions it should have, is the harder problem Pindrop is calling “agentic identity,” and BotStopper is positioned as the first piece of that larger puzzle rather than a finished solution to it.
A Real Example
Pindrop points to a deployment at a Fortune 500 healthcare organization as its clearest evidence the approach works. Bot calls to the company peaked at 6,668 in a single month, a volume that put roughly $15.1 million in HSA account balances at risk of fraudulent exposure through automated probing calls. After deploying BotStopper, that healthcare organization saw bot call volume drop by 94.3% within four months, down to a small fraction of the original peak. The attacks did not stop entirely, and Pindrop is not claiming they did. What changed is that the overwhelming majority of automated calls got flagged and handled differently than a real customer’s call would be, before any damage was done.
Why This Matters Beyond One Vendor
Pindrop’s new Chief Product Officer, Nicholas Holland, has framed the underlying challenge as three questions every business now has to answer for every interaction: is this the right entity, do they have the right intent, and are we enabling the right action. Identity verification alone does not answer any of those, which is why this problem is bigger than one company’s product roadmap.
The honest caveat here is that Pindrop has an obvious commercial interest in describing this as an urgent, large-scale problem, since selling the fix is its business. That does not make the underlying numbers wrong, the 1,680% attack increase and the IDC agent projection both come from sources outside Pindrop’s own marketing, but it is worth reading the framing with that incentive in mind.
Frequently Asked Questions
What is “agentic identity”?
It is the idea of verifying not just whether a caller is human or machine, but which specific AI agent is calling, who it represents, and what actions it should be allowed to take on a call.
Does BotStopper block all AI agent calls?
No. It is designed to detect and flag AI voice agents in real time so a business can decide how to handle them, not to block every automated call outright, since some agents are legitimate.
Is this only relevant to call centers?
The current product focuses on voice channels and contact centers specifically, though Pindrop has said it plans to extend similar verification to video and chat as AI agents spread across more communication channels.
Related: Island Raises $400 Million to Police What AI Agents Can Touch and OpenAI’s Own AI Agents Put 53 Users’ Private Photos on the Open Internet.
Bottom Line: The interesting shift here is not that fraud detection got faster, it is that businesses now have to ask a question that did not used to make sense: not just “is this person who they say they are,” but “is this even a person.” That question is not going away as AI agents get better and more common, and the companies that figure out how to answer it in under two seconds have a real edge over the ones still treating every call as if a human is on the other end.
Follow Teck Hustlers: Facebook · X · Instagram · LinkedIn · TikTok · Pinterest
