Sublime Secures $150M to Advance Agentic Email Protection

An email security startup led by a former U.S. Department of Defense cyber official raised $150 million to reduce risk and automate detection and response.

See Also: AI Agents Demand Scalable Identity Security Frameworks

Sublime Security said the Series C funding will help the Washington D.C.-based company not just block threats but also adapt rapidly to adversary behavior given the rise of AI-powered attacks, said co-founder and CEO Josh Kamdjou. He said Sublime’s autonomous security analyst and autonomous detection engineer reduce risk around email protection and make security teams more efficient.

“The problem’s getting worse,” Kamdjou told Information Security Media Group. “From a threat landscape perspective, we’re seeing adversaries adopt more and more generative AI. We want to double down on those product improvements. We’ve got a ton of investments that we want to make into the product and also expand the impact that we have on this problem globally.”

Sublime Security, founded in 2019, employs 191 people and has raised nearly $245 million, having most recently completed a $60 million Series B funding round in December 2024 led by IVP. The firm has been led since its inception by Kamdjou, who started working for the Defense Department in high school and spent 10 years engaged in offensive cyber initiatives and red teaming in the private sector (see: Proofpoint, Cloudflare Dominate Email Defense Forrester Wave).

How Sublime’s AI-Powered Agents Mimic Skilled Human Analysts

Kamdjou said the threat landscape is becoming more dangerous with the widespread use of generative AI by adversaries, who now possess greater capabilities to craft sophisticated, targeted and rapidly changing email-based attacks. The Series C investment will help Sublime build out additional agents, improve automation and scale globally to address a fast-changing threat environment, Kamdjou said.

“We have had incredible growth and momentum since we raised our Series B in terms of customer growth and ARR and overall momentum for the company,” Kamdjou said.

Rather than relying on static rules or rigid machine learning models, Sublime has built a framework of AI-powered agents that mimic the capabilities of skilled human analysts, but at machine speed and scale. Sublime’s first two agents have been deployed to detect, triage, investigate and respond to threats on a per-customer basis, adapting within hours instead of the weeks or months typically required, he said.

“We want to continue to reduce risk in the email-originating domain, and we want to continue to push the boundaries of how much time and how much more effective and efficient we can make security teams,” Kamdjou said.

Sublime uses chain-of-thought logic and a decision-making framework that helps it deal with never-before-seen attacks, which Kamdjou said is a critical differentiator. Unlike vendors that must retrain centralized ML models and deploy updates slowly, Kamdjou said Sublime’s architecture allows instant, tenant-specific updates, reducing both missed threats and legitimate messages that are incorrectly blocked.

“We use AI to better detect attacks and fight fire with fire,” Kamdjou said. “One of the big things that ASA [autonomous security analyst] does is it acts as a level one, level two analyst to investigate and triage attacks.”

How Sublime’s Detection Stack Minimizes Unnecessary Alerts

Kamdjou said Sublime’s detection stack is designed to minimize unnecessary alerts, using its agents to pre-process and investigate threats before human involvement is needed. The company’s agentic ecosystem is also specifically engineered to automate repetitive, low-leverage tasks, which Kamdjou said frees analysts to focus on the most important and nuanced security decisions.

“We have lots of tools that generate lots of alerts, and we have limited humans,” Kamdjou said. “We are trying to limit what even gets sent or shown to an analyst. We want to build agents and workflows that remove as much of the work as possible from humans so that they can focus on the things that really matter and the things that they are highest leverage for.”

Traditional email security vendors rely on centralized ML models that must be trained, regression-tested and pushed to all customers, meaning any adaptation to a specific customer’s environment could take weeks or months. Sublime avoids this entirely by delegating detection and learning to the edge, allowing per-tenant customization and adaptation, meaning updates can be made in just hours.

“You have to retrain a centralized model that gets pushed out for everybody, and it’s just a really hard problem to get right,” Kamdjou said. “What that means from a customer perspective is you’re dealing with the same attacks landing over and over, or you’re dealing with the same false positives. Sublime can just adapt to a per-customer basis. We call Sublime’s approach the distributed detection model.”

Sublime’s key performance indicators revolve around measuring impact, including quantifying how many people and mailboxes the company protects, the types of organizations it supports, and the tangible risk reduction delivered to those customers. In many environments, Sublime saves “numerous FTEs” worth of time, helping security teams reallocate resources toward higher-value activities.

“I got in security because I wanted to make the world a better place,” Kamdjou said. “I wanted to have an impact on the world, and that’s very much what drives us here today at Sublime. There’s really important national security interests that we protect, from critical infrastructure to national security systems to organizations I can’t name by name here.”