Enterprise security teams are facing a crisis they didn't anticipate: AI agents are being deployed faster than security frameworks can adapt to protect them. A new London-based startup called Trent AI just raised $13 million to address this structural gap, emerging from stealth on April 7, 2026, with a multi-agent security platform built specifically for agentic environments rather than retrofitted from traditional security tools .

The timing reveals a stark reality about enterprise AI adoption. According to a Deloitte survey cited by Trent AI, while 74% of companies plan to deploy agentic AI within two years, only 21% report having a mature governance model for autonomous agents . That's a 53-percentage-point gap between ambition and readiness, and it's creating an opening for security-focused startups to step in.

Why Are AI Agents Outpacing Security Frameworks?

The problem isn't that enterprises don't care about security. Rather, AI agents operate fundamentally differently from traditional software systems. Agents make autonomous decisions, call external tools, and execute tasks without human intervention at every step. Traditional security tools rely on static rules and predefined threat patterns, which don't translate well to systems that are, by design, unpredictable and adaptive .

Trent AI's founding team understood this challenge intimately. CEO Eno Thereska previously worked as a Distinguished Engineer at Alcion (acquired by Veeam), AWS, and Confluent. Co-founder Neil Lawrence is the DeepMind Professor of Machine Learning at the University of Cambridge and served as Director of Machine Learning at Amazon. The third co-founder, Zhenwen Dai, was a machine learning scientist at AWS and Senior Research Manager at Spotify . This combination of academic depth and operational scale at massive cloud providers gave them insight into how enterprises actually deploy AI at scale.

"The right time to build the long-term foundations of security for agentic systems," said Saul Klein, co-founder and executive chairman of Phoenix Court, the home of LocalGlobe, one of the lead investors in the round.

Saul Klein, Co-founder and Executive Chairman of Phoenix Court

How Does Trent AI's Multi-Agent Security Approach Work?

Rather than deploying a single security tool, Trent AI uses four types of specialized agents running continuously in parallel, each with a distinct role in the security lifecycle :

• Scan Agents: Observe code, infrastructure, dependencies, and runtime behavior to locate potential risks before they escalate.
• Judge Agents: Classify and prioritize security signals based on real business impact rather than relying on predefined rules that may miss context.
• Mitigate Agents: Patch vulnerabilities and validate fixes automatically, reducing the time between detection and remediation.
• Evaluate Agents: Track risk trends over time and benchmark against industry standards to measure security posture improvements.

The feedback loop between these layers is designed to improve the accuracy of each subsequent cycle. As one agent detects a threat, the others refine their understanding of what constitutes a real risk in that specific business context .

The seed round was led by LocalGlobe and Cambridge Innovation Capital, with participation from high-profile angel investors including Joaquin Quiñonero Candela, a member of technical staff at OpenAI; Avinash Bhat, a Director at AWS; Ippokratis Pandis, a Distinguished Engineer at Databricks; and Tony Jebara, former Vice President of Engineering and Head of AI and Machine Learning at Spotify . The caliber of investors signals how seriously the AI industry is taking the security gap.

What Does This Mean for the Broader AI Agent Ecosystem?

Trent AI's launch arrives at a moment when agentic AI is becoming a mainstream enterprise priority. In Q1 2026, CEO discussions about agentic AI rose 15.8% quarter-over-quarter to 6.3% of earnings calls, while mentions of AI agents more broadly jumped 27% to 6.5% of calls . Manufacturing and information and communications sectors led the charge, with physical AI (robots, drones, autonomous systems) seeing even steeper growth of 116.5% quarter-over-quarter .

The company is already working with design partners including Canopy, Commscentre, ML@Cam, Qbeast, and Weblogic, and has positioned itself as a partner member of OWASP (the Open Worldwide Application Security Project) and a startup partner with Carnegie Mellon University's CyLab Venture Network . The product also includes an open-source security agent for OpenClaw, the open-source agentic framework that has become a focal point for enterprise AI development .

What makes Trent AI's approach newsworthy isn't just the funding or the team. It's the recognition that security for AI agents requires a fundamentally different architecture than security for traditional software. As enterprises accelerate their deployment of autonomous AI systems, the companies that can help them do so safely will become essential infrastructure. Trent AI's $13 million bet suggests investors believe that moment has arrived.