As AI impersonation fraud booms, startup imper.ai has raised $28 million to stop deepfakes in real time.

One year later, this is exactly what is happening. The Identity Fraud Resource Center reports that between April 2024 and March 2025, identity fraud spiked 148%. This was caused by scammers creating fake business websites, deploying life-like AI chatbots, and generating voice agents that were indistinguishable from genuine company representatives. In 2024 alone, the Federal Trade Commission recorded $2.95 billion in losses related to identity fraud.

Now, a new startup is stepping directly into this breach. imper.ai aims to thwart AI impersonation attacks in real time, and today announced its general availability and $28 million in new funding. Redpoint Ventures and Battery Ventures led the investment round, with participation from Maple VC, Vessy VC, and Cerca Partners.

Imper.ai says that rather than trying to discover visual or audio anomalies (an approach that is quickly becoming nearly impossible), it analyzes digital breadcrumbs that attackers cannot forge. These include device telemetry (background data emitted by devices such as location, operating system, hardware details, and network activity), network diagnostics, and environmental signals. Its platform runs silently across systems such as Zoom, Teams, Slack, WhatsApp, Google Workspace, and IT help desk environments, flagging risky sessions before humans are fooled.

CEO Noam Awadish, a veteran of self-driving pioneer Mobileye and a longtime member of Israel’s 8200 Cyber ​​Warfare Squad, said AI has enhanced classic social engineering tactics, the type of attacks that manipulate people into giving up sensitive information or authorizing acts that compromise security. Through impersonation, false urgency, or psychological pressure, attackers are increasingly using AI to trick victims into revealing passwords, financial details, or remote access.

A recent example is Jaguar Land Rover. Last month, hackers used fake credentials to impersonate JLR’s IT support staff in a coordinated phishing and “vishing” (voice phishing) campaign to harvest credentials and gain access. The attack forced the automaker to shut down critical IT systems and ultimately production lines, costing the automaker an estimated $1.5 billion in losses.

Imper.ai’s founding team, Awadish, along with 8,200 other veterans Anatoly Blighovsky and Rom Dudkiewicz, believe their background as both cyber attackers and defenders gives them an edge. “I don’t think people understand that most large-scale breaches start with social engineering,” Awadish told me, adding that AI is a game-changer because email, video, and voice cloning have become nearly perfect.

Additionally, collaboration tools have proliferated far beyond email and phone calls, he noted. Attackers now have dozens of communication tools that allow AI to generate “spear phishing” messages (personalized phishing emails) at scale, as well as cloned voices and deepfake videos at high speed.

As such, imper.ai avoids attempting to detect AI impersonation directly from the AI-generated content itself. “We don’t want to get into an AI arms race,” Awadish said. Instead, the startup focuses on things that attackers cannot disguise, primarily metadata.

As the company’s momentum accelerates, so does investor interest. “We want to build a platform that protects the entire communications space,” Awadish said. “This is not a small thing. It’s not like a plugin that one of the giants is going to build.” He said the new funding will allow the company to double its research and development headcount in the U.S. and triple its go-to-market organization.

“The traction is very high at the moment so we need to keep up with the pace and we need to grow,” he said.

Note: I’m super excited to be heading to San Francisco for Fortune Brainstorming AI on Monday and Tuesday. On the main stage, we’ll be interviewing Prakhar Mehrotra, senior vice president and global head of AI at PayPal, and Marc Hamilton, vice president of solution architecture and engineering at Nvidia. I will also be moderating a spicy roundtable session on AI data centers. We’re also looking forward to meeting other speakers, including actor Joseph Gordon-Levitt, OpenAI COO Brad Lightcap, and Databricks CEO Ali Godoshi.

So, here’s more AI news for you.

sharon goldman
sharon.goldman@fortune.com
@SharonGoldman

The fate of AI

Microsoft AI wants all employees to be AI native by the end of the fiscal year, says Liz D’Anzico, vice president of design.–Written by Angelica Ahn

China’s ByteDance may be forced to sell TikTok in the US, but its quiet lead in AI could help it survive and perhaps thrive– Written by Nicholas Gordon

Anthropic considers IPO despite warning that excess liquidity is creating a bubble in the market– Written by Jim Edwards

Sam Altman declares ‘code red’ as Google’s Gemini soars — three years after ChatGPT led Google CEO Sundar Pichai to do the same– Written by Sharon Goldman

ServiceNow president says acquisition of identity and access management platform Veza will allow customers to track the whereabouts of their AI agents (by Jeremy Kahn)

AI in news

Legal AI startup Harvey raises $160 million at $8 billion valuation. Harvey, one of the fastest-growing startups in the AI ​​legal tech boom, has raised $160 million at a valuation of $8 billion, the magazine reported. new york times. The deal more than doubles the company’s valuation since February, bringing its total raised this year to about $760 million. The four-year-old company, which is already used by about half of the Am Law 100, is developing an AI assistant that helps lawyers draft and review documents, answer questions about case law, and automate routine workflows. The round was led by Andreessen Horowitz and included participation from T. Rowe Price, WndrCo, Sequoia, Kleiner Perkins and others, demonstrating that investor enthusiasm for AI tools built for white-collar professionals remains strong despite turmoil in the broader technology market.

VC “kingmaking” is happening faster than ever in AI startups. AI ERP startup DualEntry has raised $90 million in Series A at a valuation of $415 million, despite being just a year old. Lightspeed and Khosla Ventures were betting that next-generation replacements for legacy systems like Oracle NetSuite could scale quickly. But TechCrunch says the size of the round has reignited questions about “kingmaking,” an increasingly common venture capital tactic of funneling huge sums of money into a single early-stage company to build category dominance. One investor told TechCrunch that DualEntry’s ARR was only around $400,000 last summer (a figure the company disputes), but the aggressive fundraising reflects a broader shift, with venture companies picking winners faster than ever before.

Why would the AI ​​write it that way? I absolutely wanted to shout out that this (long) essay published in the New York Times was worth reading. They argue that AI-generated text, with its now-familiar combination of em dashes, ghostly metaphors, triplets, and polished sincerity, has quietly become the dominant voice of the internet, shaping everything from student essays to political speech. What’s worrying, the authors write, is not just that AI prose is everywhere, but that humans are unconsciously starting to imitate AI prose, creating a feedback loop in which machine-produced language becomes the default cultural tone. Personally, I’ve heard about how AI chatbots like the word “explore,” but that doesn’t mean they like ghostly words and all things “quiet.” “Everything is a shadow, a memory, or a whisper. They also prefer silence. For no apparent reason, and often against the logic of the story, they describe things as being quiet or quietly humming.”

Microsoft lowers sales force growth targets for new AI software. Like other Big Tech companies, Microsoft spent much of 2025 loudly touting AI agents as the next big leap in enterprise automation, but as the year ended, the company quietly dialed back expectations, a new report from The Information reveals. Microsoft eased allocations for certain AI products after several sales teams missed aggressive growth targets. This is an unusual public admission that traditional businesses remain reluctant to pay for advanced automation. Customers say ROI remains difficult to measure and the technology is too error-prone for high-risk workflows such as finance and cybersecurity. While AI has been a huge boon to Microsoft’s cloud business, thanks in large part to huge spending from OpenAI and strong demand for tools like Microsoft 365 Copilot and GitHub Copilot, getting mainstream companies to significantly increase their AI budgets has proven much more difficult than selling to AI labs.

AI calendar

December 2nd – 7th: NeurIPS, San Diego.

December 8th-9th: fortune brainstorming A.I. San Francisco. Click here to apply for participation.

January 7th to 10th: Consumer Electronics Show, Las Vegas.

From March 12th to 18th: Austin, SW.

March 16th-19th: Nvidia GTC, San Jose.

April 6th to 9th: HumanX, San Francisco.

Pay attention to AI numbers

221 million

That’s the number of YouTube users who are subscribed to so-called “AI slop” channels, channels that primarily post AI-generated content, according to a new report from Kapwing, a cloud-based video editing platform.

This report analyzed 15,000 YouTube channels from 21 countries to identify which channels are posting AI-generated content. We then looked at views, total subscribers, and estimated revenue to determine where “AI slop” channels compete most fiercely with human creators.

Some notable findings from the report: