Sovereignty and security

An AI you can present to your CIO, your CISO and your DPO.

Every commitment on this page comes from our technical documentation and our DPA: verifiable, contractual, auditable.

99.5%AvailabilitySLA over a rolling 90 days, public status page
48 hBreach notificationUpdated as the investigation progresses
10 dExport or erasureBusiness days, per person, JSON export
4 h / 24 hRTO and RPORecovery plan reviewed and tested every year
CIOArchitecture, AI and deploymentCISOAccess, encryption, audit and recoveryDPOGDPR, rights and subprocessors

Trusted by

100,000+ users satisfied across Europe
01 · Architecture

Where your data goes, component by component.

User's browser
Your applicationOne script, nothing else
Weloop widgetIsolated from your page
Your team's browser
Back officeSAML / OIDC SSO through your IdP
Scaleway, Paris region (fr-par)Data in France
Encrypted entry pointTLS 1.2+, HTTPS only
Widget, back officeBoth interfaces
API and MCP serverScoped, expiring keys
Weloop AI agentNo public access at all
Background jobsReminders, waves, backups
CacheSession data, ephemeral
DatabasePrivate network, encrypted at rest
FilesPrivate storage, time-limited access
Daily backups kept for 7 days in another EU region, and for 30 days before each deployment.
Outside the cluster
Mistral AI, EUZero retention, no training. Requests are sent, nothing is kept.
DeepL API Pro, EUTranslation, optional and can be turned off
Your identity providerOkta, Entra ID, Google, Keycloak…
Your tools, optionalJira, ServiceNow, Slack, Teams, two-way ticket sync
0 AI processing outside the EUNo third-party trackersEvery AI feature can be turned off per projectAnswers labeled as AI (AI Act, art. 50)
02 · AI processing

What the AI receives, what it keeps, what it cannot do.

Weloop keeps these results in France and deletes them along with the conversation, ticket, project or organization.

  • Transparency. Every answer is labeled as AI.
  • Review by your team. Drafts and questionnaires are reviewed before publication.
  • Limited actions. No ticket without confirmation, never a configuration change.
  • Isolation. Each processing operation stays within a single project.
FeatureData sentKept
Conversational assistantThe message, language, attachments and page contextAnswers, conversation state
Request to reach your teamThe start of the messageNothing
Article and ticket searchThe search textSearch index
Session recordingThe conversation and the recording, inputs masked by defaultSuggested values
Document importThe imported documentDrafts to review
Summary and titleThe conversationSummary, title
Release note, questionnaire, ticket fieldsThe ticket and linked conversationsDrafts, once approved
TranslationoptionalThe texts to translateCached translations
03 · Deployment

Three modes, from the cloud to your own infrastructure.

Weloop cloud

We host and run everything, in France and the EU. An access to Weloop and a script, automatic updates, 99.5% SLA.

Scaleway Paris, backups in another EU region

On your own domain

Weloop hosts the platform, your employees see it under your own domain name.

Same foundation, same SLA

Self-hosted

Helm chart for Kubernetes or Terraform for AWS ECS. Sign-in through your IdP, secret rotation. Your data never leaves your information system.

Telemetry with no personal data, manual export for air-gapped sites
04 · Security and continuity

Security as the foundation, not an option.

For your CISO's questions, the team answers at security@weloop.io.

IdentitySAML 2.0 / OIDC SSO, IdP groups mapped to roles, signed JWT or email code.
AccessServer-side RBAC. The technical operator has no access to project content.
EncryptionTLS 1.2+, OWASP Top 10, secrets encrypted with AES-256-GCM, database encrypted at rest.
AuditAnnual independent audit, critical findings fixed within 30 days.
BackupsDaily, in another EU region, kept 7 days, and 30 days before each deployment.
RecoveryPlan tested every year: RTO 4 h, RPO 24 h.
ReversibilityAgent rules versioned and reversible, data exportable at any time.
05 · GDPR

Privacy by design, all the way to session replay.

The record of processing activities, the technical measures and the data subject request procedure are in the documentation.

Data protection documentation
DPOExternal DPO registered with the CNIL (Cabinet DPO 101). DPA available from gdpr@weloop.io.
RightsJSON export or erasure of a person's data within 10 business days.
BreachNotified within 48 h, updated as the investigation progresses.
LogsRaw IP never stored, logs masked and expired after 7 days.
Session replayOff by default, consent for each recording sent, three masking modes.
AccessibilityWCAG 2.2 AA, widget in 17 languages, back office in French and English.
06 · Subprocessors

All our subprocessors, and what leaves the EU.

SubprocessorPurposeLocation
ScalewayHosting, database, filesFrance, EU backups
Mistral AIAI answers, summaries, reading attachmentsEU
Weloop AI agentFeedback assistantFrance
Bunny.netDelivery of the script and attachmentsEU
SMTP relay, LettermintSending and receiving emailEU
DeepLoptionalMachine translationEU
Microsoft Teams, Jira, ServiceNowoptionalIntegrations, once enabledEU
Slack, GoogleoptionalSlack integration, sign-in with GoogleUnited States

Published list, 30 days' notice. No AI processing outside the EU. US integrations are optional and only come into play once enabled. Self-hosted deployments choose each provider themselves.

07 · Questions

The questions from your CIO, your CISO and your DPO.

All questions
Does our data leave France?

Not for hosting: database, files and backups are with Scaleway, in France, with backups in another EU region. AI processing stays in the EU. Only the optional Slack integration and sign-in with Google go through the United States, and only if you enable them.

Does Mistral AI keep our data?

No. Zero retention and no training: requests are sent, nothing is kept by Mistral AI. Weloop keeps the results in France and deletes them along with the conversation, ticket, project or organization.

Can we limit what the AI does?

Yes. Every AI feature can be turned off per project. The agent creates no ticket without confirmation and never changes the configuration. Every answer is labeled as AI (AI Act, article 50).

Who can access our project content?

Your users, according to their roles: server-side access control, your IdP groups mapped to roles through SAML 2.0 or OIDC SSO. The technical operator has no access to project content.

What happens if there is an incident?

A breach is notified within 48 h, then updated as the investigation progresses. The recovery plan is tested every year: RTO 4 h, RPO 24 h. Backups are daily, in another EU region.

How do we get the DPA and start a security review?

The DPA is available from gdpr@weloop.io, and the team answers your CISO's questions at security@weloop.io. For the architecture, the security questionnaire or a self-hosted trial, book a slot.

Your security review, prepared with you.

Architecture, DPA, security questionnaire, self-hosted trial: book a slot with the team.

Book a slotContact us

Learn more:The productDeployment optionsFrequently asked questions