(Industries) Banking & Fintech
Kyde detects when an agent in KYC, AML or lending stops following the procedure it was approved for.
Behavioral scoring for AI agents.
For banks and fintechs that run agents in KYC, AML, lending and reporting.
Where an agent in the back office goes wrong.
(Drift)
After an update, a KYC agent approves cases it used to escalate.
(Detection)
An AML agent stops flagging a pattern it used to flag.
(Data)
An agent reads customer data that does not belong to the case.
Banks and fintechs are moving agents into the back office.
- (Back office)Agents check KYC files, prepare credit documents, monitor transactions for AML and draft regulatory reports.
- (BaFin)In December 2025, BaFin made clear that AI falls under the existing DORA risk management, with testing and the management of third-party providers.
A transaction monitoring agent closes twice as many cases overnight.
None of them is escalated. Its reports look as they always do.
(What happened)
- 22:00 A transaction monitoring agent starts its nightly alert review after a model update.
- 23:10 Normally it closes about half of its alerts and escalates the rest. Tonight it closes nine in ten.
- 06:00 Left alone, it would have closed 212 alerts without a person seeing them.
(What Kyde did)
The share of closed alerts leaves its band within the first hour. Kyde stops the agent and hands the open alerts to the AML team.
*Illustrative. Times and numbers are an example, not a customer's.
Kyde compares each run of an agent with the agent's normal behavior, using frequencies and statistics.
No language model is involved in the decision. The same record always produces the same score, and anyone with the record can recompute it.
Language models may be used later to help explain a finding. They do not make the decision. When an agent deviates, a person decides what happens next.
- DORA, ICT risk management, testing and ICT third-party risk, in force since 17 January 2025
- BaFin guidance on ICT risks in the use of AI, 18 December 2025: AI systems are covered by DORA's ICT risk management
- AMLR, Regulation (EU) 2024/1624, applies from 10 July 2027
- EU AI Act, Annex III point 5(b), creditworthiness. High-risk obligations apply from 2 December 2027
- KI-MIG, in force since 29 July 2026: BaFin supervises AI used in regulated financial activity
*Which of these applies depends on the process the agent runs. Kyde is built for them. It does not make a process compliant.
How to start
As a design partner, free of charge.
-
(Step 1) Connect
We connect to one agent you run in KYC, AML or lending. Read only, nothing changes in how it works.
-
(Step 2) Two weeks
Kyde learns the agent's normal behavior and shows you each run that deviates from it.
-
(Who you talk to)
Joerg Radehaus, co-founder, product · joerg@kyde.com Jürgen Eckel, co-founder, technology · juergen@kyde.com
*We work with a small number of design partners at a time.
(What is measured)
Kyde's detection is measured on software agents under prompt injection, in the AgentDojo benchmark. We have no measurement on a KYC or AML agent in production yet. That is what we do with design partners.
(What Kyde does not see)
Kyde does not judge whether a single case or alert was decided correctly. It sees when the share of approvals, closures and escalations leaves its usual range, and when the agent reads data outside the case.