TrustScale: the factory infographic
Measures AI against a customer's own criteria: training, evaluation and assurance.
What the company does
In normal terms
Configures and measures AI against a customer's own criteria (AI training, evaluation and assurance).
TrustScale configures and measures AI against criteria each customer defines.
In lay terms
TrustScale lets a company write its own test for its AI, and then marks the AI against it.
The risks it faces, especially from AI agents
In normal terms
- Criteria for agents. Agents act, not only answer; criteria must cover actions and access.
- Whose criteria?. Criteria written by one team may miss what security, legal or insurers need.
- Assurance over time. An agent that passed once may change.
In lay terms
- Criteria for agents. Testing what the robot says, but not what it does.
- Whose criteria?. One teacher writes the exam, others need to grade it.
- Assurance over time. Passed the test last month; what about today?
How RiskMandate helps
- Reachwhat it can actually access or affectWhat the agent can do, to test against.
- Mandatewhat the business authorisedThe criteria, written as a mandate.
- Gapthe difference: the riskWhere to aim the tests.
- Barrierswhat really stands in the wayTrustScale's assessments as evidence.
- Provenancewho owns it, who reviewed it, which versionWhich version was assessed, when.
In normal terms
Measuring an agent against its behaviour policy is assurance against your own criteria. RiskMandate's mandate can become TrustScale's criteria, and its results the policy's evidence. Met on the Policy Summit panel.
In lay terms
RiskMandate writes the exam from the business's rules; TrustScale marks it.
A first line for the conversation: You measure AI against the customer's criteria. A behaviour policy is those criteria for an agent.
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): 'Whose AI is it anyway?', 10 Nov, Policy Summit, Stage 11, names TrustScale as the speaker's organisation
- websummit.com, read 2026-11-01: LIS26 attendee page: Lawrence Snapp, CEO, TrustScale; country and industry as listed by Web Summit; public bio
- trustscale.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.