Feedzai: the factory infographic
Portugal's AI-native fraud prevention scale-up, speaking on 'Who let the agent in?'
What the company does
In normal terms
AI-native fraud and financial crime prevention platform; founded in Portugal (its CTO, the first employee, is listed by Web Summit from Portugal).
Feedzai uses AI to prevent fraud and financial crime; founded in Portugal.
In lay terms
Feedzai uses AI to spot fraud, such as fake or stolen payments, for businesses that handle money.
The risks it faces, especially from AI agents
In normal terms
- Agents that pay. Shopping and payment agents act for customers; a fraudulent agent and a legitimate one look alike.
- Who authorised it?. Fraud decisions need to know what the customer actually authorised the agent to do.
- Agents inside banks. Banks' own agents in operations and SaaS reach sensitive data and payment flows.
In lay terms
- Agents that pay. A robot is buying things with your card. Is it yours?
- Who authorised it?. Did you tell your assistant to spend that much?
- Agents inside banks. The bank's own robots can move money too.
How RiskMandate helps
- Reachwhat it can actually access or affectWhat a payment agent can spend and where.
- Mandatewhat the business authorisedWhat the customer authorised.
- Gapthe difference: the riskSpending beyond the authorisation: a fraud signal.
- Barrierswhat really stands in the wayLimits, step-up checks, Feedzai's models.
- Provenancewho owns it, who reviewed it, which versionWhose agent, which version, approved when.
In normal terms
A stated mandate per agent is a strong fraud signal: an agent acting outside its published policy is suspicious. For Feedzai it is a new input; for its bank customers, a governance record. A local Portuguese anchor for the agent-authorisation story.
In lay terms
If every shopping robot carried a note saying what its owner allowed, spotting the fake ones gets easier.
A first line for the conversation: 'Who let the agent in?' Our answer: the agent's own policy, saying what it was allowed 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): 'Who let the agent in?', 12 Nov, SaaS Summit, Stage 7, names Feedzai as the speaker's organisation
- websummit.com, read 2026-11-01: LIS26 attendee page: Diogo Guerra, CTO, Feedzai; country and industry as listed by Web Summit; public bio
- www.feedzai.com, 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.