Paul is built for organisations that take data seriously. EU hosting, strict no-training policy, organisation-level isolation — here are the concrete commitments you can audit.
Our entire infrastructure runs within the European Union:
No customer data leaves the EU for application processing. Inference requests to model providers are made with explicit training opt-out flags on the provider side.
We never reuse your templates, prompts, presentations or metadata to train or fine-tune any model, whether internal or operated by a partner.
All calls to partner LLMs (OpenAI, Anthropic, Mistral) are made with data_retention and training_opt_out flags enabled. Request logs are retained strictly for billing and observability purposes, then purged according to our retention policy.
Each customer organisation has a dedicated isolated space enforced at the database level via Postgres Row Level Security. No query can cross organisation boundaries — by design.
Roles (administrator, editor, external consultant) are managed granularly and audited. A user removed from an organisation loses access immediately, with no residual access.
An internal security audit covering 13 cybersecurity domains was conducted in April 2026; priority fixes were applied across all environments.
Paul is published by Logitopia SAS, a French legal entity subject to GDPR and French law.
These features are available on request as part of an Enterprise engagement. Let's discuss your requirements during a personalised demo.
We respond individually to every RFP, security questionnaire, or DPA. Book a demo — we will take the time to understand your requirements.