Regulators do not ask whether an organisation uses AI agents. They ask who approved a change, what data the system used, whether a human could intervene, and whether the organisation can reconstruct what happened. ATX Agent Hub is built so that those answers exist as a by-product of running the agent — not as a project after the fact.
No software makes an organisation compliant on its own; compliance comes from your processes, your risk assessment and your use of the tools. What a platform can do is provide controls that produce evidence. The mapping below shows which ATX Agent Hub controls support which obligations, so your compliance and risk teams can make their own assessment. It is not legal advice.
The EU AI Act asks providers and deployers of higher-risk AI systems for human oversight, record-keeping, transparency, accuracy and robustness. Several of those expectations correspond directly to platform features. After the 2026 Digital Omnibus, the high-risk obligations apply from 2 December 2027 for the use cases listed in Annex III, such as credit scoring.
The Digital Operational Resilience Act puts ICT change management, logging, access control and third-party risk on the agenda of every EU financial entity. AI agents are ICT systems like any other, and a self-hosted deployment keeps them inside the entity's own resilience framework.
Personal data in documents and prompts is personal data processing. Project isolation, per-source access control in knowledge bases and self-hosted deployment help keep processing within a defined purpose and location, and retention policies are set centrally in the Admin Panel.
Banks increasingly treat generative AI under model-risk frameworks: inventory, validation, monitoring and change control. Workflows in ATX Agent Hub are versioned inventory items with owners; tests provide validation evidence; run history, cost tracking and regression dashboards provide monitoring; and the model behind each step is recorded on every run.
Each production workflow can carry a manual — a version-controlled SOP or runbook pinned to the workflow and environment, approved before publication. The description of what an AI agent is meant to do lives next to the agent itself, and changes to it go through review.
No software does that on its own. The platform provides controls — human oversight, logging, transparency, testing, change control — that support the obligations, and produces the evidence your compliance team needs for its own assessment.
Through change management with immutable versions and independent approval, role-based access, audit logging and replayable runs, and through self-hosted deployment that keeps AI agents within the entity's own ICT environment.
Yes. Each run's trace records the inputs, the knowledge-base passages retrieved, the model used, every step's output and the final answer with its citations.
This site needs JavaScript for the full experience. Every page above is reachable without it.