Trent AI
Sources: Funding announcement (BusinessWire) · SecurityWeek coverage · Tech.eu coverage
Homepage URL above is unverified — Trent AI’s primary domain has not been disclosed in coverage as of 2026-05-03; re-validate before external citation.
What
London-based agentic AI security startup founded in 2025 by former AWS engineering leaders, led by CEO Eno Thereska. Builds a layered platform meant to secure AI agents throughout their entire lifecycle — pitched at developers and organizations building AI agents and autonomous software systems (Source: SecurityWeek).
Funding
$13M (€11M) seed round, April 7, 2026 — led by LocalGlobe and Cambridge Innovation Capital, with backing from leaders at OpenAI, Spotify, Databricks, and AWS as angels.
Second-largest agentic-AI-security seed round in the May 2025–May 2026 window, behind only Lumia.
Relevance
Cross-plane coverage: the multiple-coordinated-agents architecture matches the Sentinels-and-Operatives split, and the lifecycle scope (“scans models, dependencies, infrastructure, runtime behavior”) spans CMM D4 (Runtime & Guardrails), D7 (Observability & Detection), and D8 (Supply Chain & AI-BOM), with D1 (Governance) as the wrapping layer.
Functionally a competitor to Onyx Platform and Lumia in the Guardian Agent / AI Agent Management Platform category. Differentiator: explicit multi-agent security architecture (“agents securing agents”), an instance of the Guardian Agent Metagovernance question — who guards the guardians.
Product
Per the launch announcement, the platform:
- Continuously scans models to observe code, dependencies, infrastructure, and runtime behavior
- Analyzes risks and business impact
- Patches vulnerabilities and modifies configurations
- Validates fixes (loop closure)
- Embeds the security layer into developer workflows
Multiple coordinated security agents work together continuously to understand the environment and secure it (Source: Trent AI press release).
Gap
No technical disclosures as of May 2026 on whether the coordinated agents are LLM-driven (vulnerable to recursive prompt injection like any LLM-as-judge layer) or rule-based. Re-validate when more material lands.
Notable Statements
- Thereska: “Organizations are deploying AI agents and autonomous workflows faster than their security can adapt” — frames the gap that the AI Agent Layered Council and Shadow Automation concepts already track.