Built for compliance from day one.
Thaliq is the agentic layer your business runs on, which means every conversation, tool call and HITL decision moves through our infrastructure. This page is the honest answer to "what happens to our data when we use Thaliq."
Where we stand on certifications.
Status reflects today, not the marketing version. When something flips state, this page updates the same day.
Four pillars under every Thaliq tenant.
We optimised the platform for multi-tenant by default, not as a feature. Here's how that shows up in your security posture.
What we store. What we never store.
The shortest version of our security story. If you only read one section, read this one — it tells your legal team what they actually need to know.
-
✓
Conversation messagesPer tenant, encrypted at rest. Retention follows your plan.
-
✓
Agent specificationsVersioned snapshots so you can roll back. Stored per tenant.
-
✓
Audit logsWho · what · when · cost — for compliance and forensics.
-
✓
Tool call metadataName, duration, success/failure, model used. Not payload contents.
-
✓
RAG documentsOnly the documents you explicitly upload to your agent.
-
×
Customer API tokensPassed through to tools on each call; never persisted on our side.
-
×
LLM provider keysOn BYOK, the key lives in your secret manager. We never see it.
-
×
Raw tool payloadsWe log the shape of a tool call, not the contents. Your end-user PII stays in your systems.
-
×
End-user passwordsAuthentication happens at your IdP. We only see the resulting JWT claims.
-
×
Conversations from churned tenantsHard-deleted 30 days after offboarding. No long-tail data hoarding.
Who else touches your data.
Honest list of the third parties that process tenant data on our behalf. Updated whenever a subprocessor changes. Enterprise tenants get notified 30 days before any change.
BYOK customers can route inference to their own Anthropic / Bedrock account, removing Anthropic as a Thaliq subprocessor for their tenant.
Legal and compliance artefacts.
We have these on hand and can send them via NDA. We do not publish them in full to keep clauses fresh and to control distribution.
Found something? We want to know.
We don't run a public bug bounty yet, but we respond to responsible disclosure quickly. Reach out to the address below; we will acknowledge within 24 hours and coordinate a fix window.
- · We acknowledge critical reports within 24 hours.
- · Standard disclosure window is 90 days, negotiable based on severity.
- · No legal action against good-faith research that follows this policy.
- · Hall-of-fame on this page for accepted reports (with your consent).
Need to deep-dive before signing?
We will get on a call with your security team, walk through controls, share what we can under NDA, and answer the awkward questions. No marketing, no slides — engineers who can answer.