Trent AI Raises $13M to Rebuild Security for the Agentic World
We founded Trent to answer a simple question: is what I’m building, whether it’s an app, a system, or an agent, actually safe? We spend hundreds of billions every year on security, yet that question is still hard to answer. We want to fix that.
Why this problem exists
Across startups, mid-sized companies, and FAANG, I kept seeing the same pattern: the relationship between security and developers was fundamentally broken. At large companies, even ones with world-class security teams like AWS, security often showed up after the code was written, forcing developers to stop, context switch, and fix problems they didn’t create. Reviews delayed launches, created friction, and often felt disconnected from how systems were actually built. With today’s agentic systems, this disconnect is even worse.
At smaller startups, the problem looked different, but just as serious. Most teams simply couldn’t afford dedicated security engineers, but the stakes were just as high. For example, at Alcion.ai (acq. by Veeam), I worked on a backup system designed to protect against ransomware. Before we wrote a single line of code, a few of us did a full threat model of the system, thinking through how an attacker might try to delete backups or corrupt recovery paths. That work mattered, it shaped the architecture in ways that made the system fundamentally safer. But it also highlighted something uncomfortable, this kind of thinking just doesn’t scale beyond the initial one-off effort. It depended on individual effort, experience, and passion for security, not on a system. And in a world where agents are generating systems continuously, relying on one-off human reasoning breaks down even faster.
At the same time, I grew increasingly frustrated with what I can only describe as (expensive) security theatre. We used tools that scanned code and surfaced hundreds of issues, 400+ findings at times. The majority were false positives or low-signal noise. They looked good in dashboards, managers tracked them and leadership wanted them to trend down, but engineers, the actual builders, largely ignored them. Why? Because, the real risks weren’t in those lists. And in a world where agents are building and operating systems continuously, managing security through dashboards simply doesn’t scale.
Risks to a business don’t come from a missing input validation, vulnerable dependencies or SQL injections. They come from decisions made much earlier, at the design and architecture level, and from the context in which a system actually operates. The same piece of code can be safe in one system and critical in another, depending on how data flows, what’s exposed, and what assumptions are made. In agentic systems, that context becomes even more dynamic, agents deciding what to access, what actions to take, and how systems connect in real time. For example, if your storage system is designed such that an attacker can delete backups entirely, no amount of code scanning will save you. That’s a design flaw, not a bug.
The Ground Shift and the Trent AI Opportunity
Meanwhile, the ground kept shifting. I watched multiple security teams struggle through transitions, from on-prem to cloud, from monoliths to microservices. Each shift required rethinking everything, often rebuilding teams, processes, and tools from scratch.
Now we’re in the middle of another shift, and it’s bigger. AI is changing how software is built. Code is generated faster, systems are more dynamic, and increasingly, agents are making decisions and taking actions with limited human oversight.
At Trent AI, we’re seeing this firsthand. The pace of development is accelerating dramatically, agents are being deployed into production workflows, and the ratio of things happening to humans reviewing them is shrinking.
All the old problems still exist: security disconnected from development, design flaws missed early, tooling producing noise instead of insight, teams overwhelmed by constant change, but now they’re amplified, with fewer humans in the loop to catch issues.
That’s why we started Trent AI. Security needs to be rebuilt for an agentic world. It needs to understand agents in context, reason about design decisions, and keep up with how software is actually being built, continuously learning as it goes. Ultimately, security should assess risk in a way that’s specific to each system and help close the gaps. Developers should be able to build agents, while systems like Trent handle the security reasoning behind the scenes.
Working in security, you end up holding a few seemingly contradictory things at the same time. Developers don’t care about security, and honestly, they shouldn’t. Their job is to build. At the same time, trust and security still matter deeply, and failures there are existential for a company.
The same tension is playing out with AI. A lot of security knowledge is being captured in models, and tools are already getting good at fixing low-level issues automatically. But as more software is generated and more agents are deployed, the attack surface is expanding just as quickly.
That tension is where the real problem sits.
Building Trent
At Trent AI, we’ve been working closely with the AI-native community, because that’s where the patterns are emerging first. The best way to learn right now is from the teams building at the frontier. Some early examples include:
We’ve been lucky that people with deep experience in AI and systems trusted us early enough to join us. The Trent AI team knows how to get AI-based products to production and delight customers. Just as importantly, this team brings the grit and long-term mindset that security demands. This isn’t a space for shortcuts or quick wins, it requires trust, consistency, and a willingness to do the hard work over time.
This $13M gives us the opportunity to go deeper on that mission. We’re lucky to have investors who were operators themselves and believe in building this the right way. The core idea is simple: security shouldn’t slow builders down, it shouldn’t rely on noise, and it shouldn’t start after the system is already built. It should be part of how systems and agents are designed, especially in a world where software is no longer written just by humans.
We’re excited to build that future.
Eno
Founder & CEO, Trent AI
