For Kevin Tian, the best solution for artificial intelligence-powered fraud is also the most obvious: pit AI against AI.
In 2022, he founded social engineering defense startup Doppel to do just that. And as cybercriminals harness ever more advanced AI models to turbocharge their attacks, Doppel’s AI systems have helped businesses combat them at scale— more quickly and effectively than before.
The startup has built AI agents— software that is programmed to autonomously carry out specific tasks— to scour the internet, the dark web and social media for potentially fraudulent activity, flagging everything from copycat websites and fake user accounts to malicious advertisements on Google, Instagram, and YouTube. Doppel’s agents screen 100 million alerts of such phishing threats every day, filtering real threats from benign ones and reporting them to platforms to be removed. Tian says they do this with about 90% accuracy, and they’re constantly improving.
“If threat actors can use AI to spin up these attacks for just a few cents on the dollar, we’ve got to make sure that we can also handle that volume on our side,” Tian told Forbes.
On Friday, Doppel announced $35 million in new funding in a round led by Bessemer Venture Partners. With $55.5 million in venture backing, it’s now valued at $205 million. It’s another milestone for the company Tian cofounded in 2022 with CTO Rahul Madduluri, whom he had met at Uber while working on the company’s flying car moonshot, Uber Elevate. Initially, the duo intended to combat NFT-related fraud, helping crypto companies track and report counterfeits. But in 2023 Doppel broadened its customer base to other industries.
In its early days, Doppel used contractors in countries like the Philippines and India to sort through thousands of potential threats and decide which ones were malicious. But in September 2024, it found that OpenAI’s new models, those capable of “reasoning,” could perform the same tasks. It replaced those contract workers with a cohort of new AI agents and used them to automate 30% of its security operations. Tian claims AI agents have been able to identify more threats than humans. It’s been transformative for the company’s business.
“It’s just not scalable to have a human team review those millions of alerts every single day and we now have an AI agent that can make those decisions,” Tian said.
In 2023, productivity software Notion was struggling with an onslaught of attacks, including malicious ads targeting its customers and efforts on social media to impersonate its CEO Ivan Zhao. In one scenario, “[fraudsters] would take our download page for our application, scrape it, make a clone of it, make a domain that looks similar. And so when you visit it, it looks like you’re visiting Notion,” said Daniel Pyykonen, head of platform security at Notion.
So it turned to Doppel for help. The company’s AI agents made quick work of those campaigns, taking down thousands. Eventually Notion started seeing fewer and fewer attacks. “We became a very expensive target,” Pyykonen said.
Doopel’s secret sauce is a “threat graph,” a map of relationships and interactions between social engineering campaigns — telephone numbers, IP addresses, advertiser accounts. It allows the company to better track malicious hackers using AI as a productivity tool and help safeguard businesses against future attacks. “We want to give the good guys that same view so they could play less whack-a-mole and shut down the entire threat actor infrastructure,” Tian said.
To date, Doppel has attracted some 180 enterprise customers including United Airlines, OpenAI and Coinbase and tripled revenue year over year, Tian claims. It has also onboarded customers in sectors like oil and gas, finance and insurance, for which it has rolled out a second set of industry-specific AI agents. “We were blown away by the breadth of tier one clients they had outside of tech,” said Elliott Robinson, partner at Bessemer Venture Partners.
Artificial intelligence has greased the wheels for cyber criminals to spin up new types of social engineering attacks that extend far beyond phishing emails embedded with dangerous links. They now include tactics like a WhatsApp message from an account impersonating a company’s CEO urgently asking for funds, a fake LinkedIn profile mining for private information and even a realistic decoy of a company’s website. In 2024, Google suspended 40 million malicious advertiser accounts from its platform. Many of them featured AI-generated images and audio of public figures and CEOs to deceive people and promote scams. Between April 2024 and April 2025, Microsoft stopped fraud attempts that would cost its customers $4 billion and blocked 1.6 million bot signup attempts on its Azure platform.
Tian likens today’s dilemma to the one Tom Cruise faced in Mission: Impossible — Dead Reckoning: a rogue AI system on a mission to distort reality and control the world. “In that movie, AI is the bad guy, it can impersonate anyone, create fake news, manipulate entire societies just through fake digital deception,” he said. He views Doppel as an opportunity to “build a company solving one of the biggest problems in the world today, which is AI enabled social engineering.”
