Meet us at Black Hat 2026 🌸

Announcements

Itay Keren
|
|
Reading Time:
3 minutes
min
For most of the last decade, the endpoint was treated as solved. The category matured, the vendors consolidated, and the assumption held: whatever runs on a laptop, someone is already watching it. That was true when the risk was a malicious executable. It stopped being true the moment the endpoint stopped looking like a set of approved programs and started looking like an ecosystem.
AI agents brought their own MCP servers. IDEs shipped their own package managers. Browsers became platforms. The endpoint stopped being a fixed set of installed programs and became something assembled on the fly: a mesh of unvetted agents, task-specific extensions, and packages pulled in on demand - most of it never reviewed, much of it never seen by security.
One pattern kept surfacing - and it's now everywhere on the endpoint.
Start with AI. Agents arrived with privileged access to data, credentials, and the ability to act on a user's behalf - pieced together on the fly from the skills and MCP servers they pull in. Adoption is moving faster than any policy can follow. It should. The work is real. But security teams are left without the basics: which agents are running, what data they touch, what they can execute, and what they actually did.
Then there's the risk that doesn't look like a program at all: extensions, plugins, packages. A few are outright malicious. Lots of them aren't. They're fine on one endpoint and present a real risk on the next, depending on who's running them, what they can reach, and what's running alongside. That's the judgment EDR was never built to make.
These don't sit side by side - they feed each other. An agent is only as safe as the skills and MCP servers it's built from. An unvetted package is one agent away from reaching everything that agent can touch. Same endpoint, each one raising what the other can do. AI didn't create this problem - it accelerated it. You can't secure one without the other. So here's the layer where the risk now lives - the agents, the extensions, the packages, the code assembling itself on every endpoint. And quite frankly, no one is watching it.
Existing tools weren't built for this. EDR solves a real problem - just a different one. It answers one question: is this thing known-bad. It can't answer the one that matters now - is this safe here, on this endpoint, for this person. That's a judgment, not a lookup.
That gap is why we created Bloom.
We'd seen it forming from the inside. The founding team spent years building enterprise security at Palo Alto Networks and through the Demisto and Dig Security acquisitions - leading the security research and building the data and AI posture products that the current generation of enterprises runs on. We understood how these tools reached the endpoint, and how little stood between them and the data once they did. We set out to close that gap for good.
Bloom is endpoint security, rethought for what the endpoint became: one platform to control, defend, and manage everything running across the fleet. It starts by seeing all of it - operating system, local data, applications, extensions, and the AI agents and code now layered on top. Then it reasons about each one the way a good analyst would - agentic, contextual, per endpoint - scoring risk against the user's role, their access, and everything else on the machine, rather than checking it against a static list. It fixes what creates risk without breaking how people work. And it prevents the next risky component from landing at all, with a supply-chain firewall that stops software at the door rather than after it's running.
One line of control. From the moment a tool tries to enter to the moment it's live on the endpoint: deciding what belongs, fixing risky configuration, blocking what shouldn't be there, and enforcing policy across the fleet. So security teams stop stitching together a partial picture from products that were never meant to talk to each other, and finally have real control over what runs on the machine.
We're grateful to the early customers who shaped this product from day one, our investors, our partners, and the amazing team that chose to join us.