Trent AI launches maturity model
- Trent AI said on May 5 it launched an AI Security Maturity Model, a scoring framework for companies building and deploying autonomous AI agents. - The model grades organizations across six domains and 28 categories, from “Partial” to “Adaptive,” to turn broad governance rules into concrete gaps. - It lands as enterprises rush into agentic AI, while far fewer have mature governance for autonomous systems than deployment plans.
AI security is starting to split into two jobs. One is blocking obvious attacks. The other is figuring out whether autonomous systems are being run with anything like adult supervision. Trent AI’s news sits squarely in that second bucket. On May 5, the London startup launched an AI Security Maturity Model meant to score how ready an organization is to secure “agentic” development — basically, software that can take actions, not just answer prompts. ### What is Trent AI actually launching? It’s not a new firewall, scanner, or model. It’s an assessment framework — the AI Security Maturity Model, or ASMM — that gives security teams a structured way to measure how prepared they are for AI systems that act on their own inside development and enterprise workflows. Trent says it analyzed prior incidents to expose the gaps. ### Why does “agentic” change the security problem? A chatbot can say something dumb and stop there. An agent can read code, call tools, touch infrastructure, trigger workflows, and chain decisions together. That means the security question is no longer just “is the model safe?” It becomes “who can invoke this thing, what can it do? Old security programs often don’t see those risks clearly enough yet. ### How is the model structured? The framework mirrors familiar cybersecurity language, but adapts it for AI agents. It scores organizations across six domains — Govern, Identify, Protect, Detect, Respond, and Recover. Inside those are 28 categories, and each category gets a maturity score from 1, labeled “Partial,” to 4, labeled “Adaptive.” The output is supposed to show both an overall score and the specific domains where a team is weak. ### Why anchor it to NIST and the EU AI Act? Because most big companies do not want another standalone checklist. Trent says ASMM is built on NIST Cybersecurity Framework 2.0 and aligned with the NIST AI Risk Management Framework and the EU AI Act. That matters because governance teams already speak in those frameworks, but expect controls. ### Why launch this now? The timing is the whole story. Trent points to Deloitte’s 2026 State of AI numbers: 74% of companies plan to deploy agentic AI within two years, but only 21% say they have a mature governance model for autonomous agents. That gap is the opening. Companies are moving from experiments to systems that can actually do things, and the governance layer is lagging badly. ### Is this bigger than one product launch? Yes — because the market is shifting from “can we build agents?” to “can we control them?” Trent only emerged from stealth a month ago with a $13 million seed round and a broader pitch around multi-agent security. This maturity model extends that pitch: not just protecting agents at runtime, but giving enterprises a way to justify budget and sequence the work before rollout gets messy. ### What’s the catch? A maturity model is only useful if teams actually map it to real telemetry and real controls. Plenty of frameworks look tidy in slides and die in implementation. Trent’s bet is that security buyers now want exactly this kind of translation layer — something concrete enough to score, compare, and fund — because the agent boom is moving faster than policy documents can. ### Bottom line This launch matters less as a standalone framework than as a signal. Agentic AI security is becoming its own category, and the winning tools will probably be the ones that turn abstract governance into operational decisions teams can act on.
