AegisAI Raises $36M to Fight AI-Generated Spear Phishing

The old phishing email usually gave itself away.

The sender looked wrong. The language felt unnatural. A suspicious link or attachment triggered a rule that security software already understood.

Generative AI is removing those clues.

San Francisco-based AegisAI has raised $36 million in Series A funding to build email defenses for attacks that can research their targets, imitate legitimate communication and generate a different lure for every recipient. Battery Ventures led the round, while existing investors Accel and Foundation Capital returned. The financing brings AegisAI’s total funding to $49 million, less than a year after its public launch.

Founded by former Google security leaders Cy Khormaee and Ryan Luo, AegisAI draws on the team’s experience working on reCAPTCHA, Safe Browsing and Web Risk-systems built to identify abuse across billions of users.

The attacker no longer needs a reusable template

Traditional email filters perform well when an attack contains something previously identified as dangerous: a known domain, malicious file signature, suspicious phrase or repeated campaign pattern.

AI-generated spear phishing can avoid that repetition.

An attacker can collect public information about an executive, supplier or finance employee and use a language model to produce a credible message built around that person’s role and relationships. The email may arrive from trusted infrastructure, contain polished language and avoid obvious malicious links until later in the interaction.

Research has already shown that automated AI-generated spear-phishing emails can perform at levels comparable with messages written by human experts while reducing the cost of personalization at scale.

AegisAI’s response is to analyze what the message is trying to make the recipient do-not only whether it matches a list of known threats.

Several agents inspect one message from different angles

The company uses a coordinated group of autonomous agents to examine sender identity, language, attachments, links, QR codes and behavioral context.

Instead of depending on one model or a long set of manually maintained rules, these agents work independently and combine their findings to determine whether a message represents phishing, business email compromise or another emerging attack. AegisAI says this approach can reduce false positives by as much as 90% compared with conventional systems, although that figure remains company-reported.

The platform connects directly to Microsoft 365 and Google Workspace through their APIs. That allows companies to deploy it without changing their mail-routing records or placing a new gateway in front of every message.

Vanguard follows the threat after it leaves the inbox

Part of the new funding will accelerate the general availability of Vanguard, an agent designed to investigate suspicious content beyond the email itself.

When a message contains a questionable link or attachment, Vanguard can follow the path an employee might take across external websites, including pages protected by CAPTCHAs, cloaked destinations and weaponized documents. It then returns a threat report intended to explain what the message was attempting to trigger.

That capability addresses a growing weakness in email filtering: the malicious behavior may not appear until the recipient leaves the inbox.

A familiar security category is becoming unsettled again

Email security has long been dominated by established vendors, gateway products and employee-awareness training.

AegisAI is arguing that those layers were designed for a period when attacks were slower to create, easier to reuse and more likely to contain recognizable warning signs.

The company says it has deployed with dozens of customers across technology and financial services, including Mesh, LangChain and Lokker. The Series A will support the expansion of its agent fleet, Vanguard’s rollout and a larger enterprise go-to-market operation.

Its toughest benchmark will not be catching obvious spam. It will be identifying the message that looks completely ordinary because an AI system designed it for one person, one company and one moment.

Source : TechCrunch