Privacy Notice

Last updated 2 September 2026

This notice explains what Risicare collects when you use the platform, how long it is kept, who it is shared with, and what you can ask us to do with it. Risicare is an observability platform for AI agents: the data we hold is mostly telemetry your own software sends us.

The controller of this data is [TBD-LEGAL: registered entity name], registered at [TBD-LEGAL: registered address]. Data protection enquiries: [TBD-LEGAL: DPO / responsible contact].

1. What we collect

Agent telemetry (spans and traces). When you instrument an agent with our SDK we receive one record per operation: identifiers, timings, status, the model and provider used, token counts and cost, plus your own service and environment names.

Prompt and completion content. The text of your LLM calls — prompts, completions, tool inputs and tool outputs — is captured only when content capture is on. Both SDKs default it off: you must set trace_content=True (Python) or traceContent: true (JavaScript) to send it. When it is sent, it is stored on the span record itself.

Account data. If you sign in with GitHub or Google we store the email address, name and avatar URL those providers return.

Operational logs. Administrative actions are recorded in an audit log that includes the originating IP address and browser user-agent string.

Contact form. If you write to us through this website we collect the name, email address, company, region and message you provide, together with the IP address the request arrived from. That message is delivered to us by email.

2. Redaction

Before telemetry is written to storage we run an automatic redaction pass over it. Credential-shaped values (tokens, keys, authorization headers) are redacted everywhere. Personal-data patterns — email addresses, phone numbers, national identifiers, payment card numbers and IP addresses — are redacted from content fields.

Redaction is pattern-based and always on; it is not a setting you can turn off, and it is not a guarantee. It reduces exposure, it does not eliminate it. If your prompts carry personal or confidential data you do not want us to hold, turn content capture off.

One bound you should know: a copy of each span is written to our object archive at the moment it is received, and the credential redaction is applied there, but the personal-data redaction pass runs later in the pipeline. Personal-data patterns can therefore be present in the archive copy even where they have been removed from the analytical store.

3. How long we keep it

Retention is enforced by fixed, table-level expiry in our data store. It applies to every project alike — it does not vary by plan, and it is not configurable per project.

• Traces, spans, sessions, agent spans and agent messages — 90 days
• Prompt and completion content — 90 days (it is stored on the span)
• Evaluations and scorer results — 365 days
• The object-store archive copy — lifecycle expiry configured at 90 days
• Account records and audit logs — retained for the life of the account. Audit logs are not currently on an automatic expiry schedule.

Telemetry passes through an in-memory queue on its way into storage. Records there are transient and are consumed within seconds under normal operation.

4. Who we share it with

We use the following processors. We do not sell your data and we do not use it to train models.

• Amazon Web Services — infrastructure hosting for all platform data.
• Together AI — the large-language-model provider used by our diagnosis pipeline. See the disclosure immediately below.
• Resend — delivery of the email generated by this website's contact form.

About the Together AI transfer. When a diagnosis runs, the failing trace's error message and stack trace are sent to Together AI together with the surrounding spans — which, where content capture is on, includes your prompts, completions, tool inputs and tool outputs. Two things bound this today. First, automatic diagnosis is off by default and must be enabled per project, and the manual diagnosis endpoint is currently disabled for the beta — so this transfer does not occur unless and until you turn it on and we re-enable the pipeline. Second, we do not currently have a signed data processing agreement with Together AI. We are resolving that, and we are telling you rather than implying otherwise.

5. Your rights

You can ask us to delete telemetry associated with a given subject — a session, an agent, or a set of traces. Write to [TBD-LEGAL: privacy contact address] and we will run the deletion for you; there is no self-service control for it yet.

You should know the scope precisely. The deletion covers the telemetry stores: spans, traces, sessions, agent records, content, evaluations, scorer results, archived copies, and the diagnosis and fix records derived from them. It does not delete your account, your organisation, API keys, webhooks, alert rules, audit logs or project configuration. Backups taken before a deletion are out of scope: restoring one reinstates the data, and we do not currently re-apply deletions after a restore.

We are also working through a known limitation in which a deletion can complete while leaving orphaned copies in the object archive. Until that is closed we cannot represent archive deletion as exhaustive, and we would rather say so than assert otherwise.

You can request a copy of your project's data at the same address. Account closure is handled by us on request rather than through a button in the product.

6. Security

Data is encrypted in transit. Database volumes use provider-managed volume encryption and the object archive uses server-side encryption with a managed key. API keys are stored as SHA-256 hashes, never in plaintext.

The bound on that: volume-level encryption protects against disk theft and snapshot leakage. It does not protect data in use on a running host, and we do not currently run database-level transparent encryption.

7. Cookies

The application sets cookies required to keep you signed in: a session cookie and an authentication cookie scoped to our domain. We do not use advertising cookies, and we do not load third-party analytics or tracking scripts.

8. Changes

If we change this notice materially we will update the date at the top and, where the change affects how we handle data you have already sent us, tell account holders directly.

9. Contact

Privacy enquiries and rights requests: [TBD-LEGAL: privacy contact address]. Security reports: [TBD-LEGAL: security contact address].

Back to risicare.ai