Trent AI: the factory infographic
An AI security engineer that finds the real risk in code, cloud and AI agents, and verifies each fix.
What the company does
In normal terms
Agentic AI security platform that works like an AI security engineer: finds the real risk in code, cloud and AI agents and verifies every fix.
Trent AI works like an AI security engineer: it finds the real risk in code, cloud and agents and checks that fixes work.
In lay terms
Trent AI is a robot security guard that walks the building, finds the weak spots, and checks they got fixed.
The risks it faces, especially from AI agents
In normal terms
- Finding risk needs intent. Judging an agent's risk needs to know what it was meant to do.
- The scanner's own reach. A security agent needs broad access, making it a high-value target.
- Fixes that drift. Fixes verified today drift as agents change.
In lay terms
- Finding risk needs intent. The guard can't tell if the open door is a problem unless it knows who should use it.
- The scanner's own reach. The guard has every key.
- Fixes that drift. The lock was fixed, then someone changed it again.
How RiskMandate helps
- Reachwhat it can actually access or affectWhat each scanned agent can touch.
- Mandatewhat the business authorisedIts intended job: the context for judging risk.
- Gapthe difference: the riskThe findings that matter most.
- Barrierswhat really stands in the wayVerified fixes as evidence.
- Provenancewho owns it, who reviewed it, which versionVersion history, so drift shows.
In normal terms
A behaviour policy gives Trent's findings their context (reach versus mandate), and Trent's verified fixes become evidence in the policy. Its own agent needs a policy too.
In lay terms
RiskMandate tells the security robot what each door is for, so it knows which open doors matter.
A first line for the conversation: Your platform finds agent risk. A behaviour policy tells it what each agent is meant to do.
Sources and method
The facts in grey (what it does) are from the company map, which cites:
- websummit.com, read 2026-11-01: LIS26 session page (conferenceSlug lis26): 'Viable systems, judgement, and AI safety', 11 Nov, AI Summit, Stage 13, names Trent AI as the speaker's organisation
- websummit.com, read 2026-11-01: LIS26 attendee page: Neil Lawrence, Co-founder & Chief Scientist, Trent AI; country and industry as listed by Web Summit; public bio
- trent.ai, read 2026-11-01: what the company does, from its own site (title and description)
The risks and how RiskMandate helps are our analysis, written by dev@web-summit.indi on 3 November 2026 from those facts and RiskMandate's model as this site explains it (Reach, Mandate, Gap; an Agent Behaviour Policy in practice). They are not the company's own statements and name no incident. A person reviews each before it is used.