AI providers
Choose the AI operating model your lab can approve
Researcher Center separates the knowledge workflow from the inference provider. Labs can use managed AI, bring an approved provider key, or run local/self-hosted models.
Source-grounded answersWorkspace boundariesDeployment choice
Who reviews this page
PIs, lab administrators, research IT, and technical reviewers deciding how AI should run for a lab corpus.
Three operating models
- Managed AI: the fastest path for an approved pilot with platform-managed inference and usage controls.
- Bring your own key: use an institutional or lab-owned provider account and keep billing under your governance.
- Local/self-hosted: keep sensitive documents and model calls inside infrastructure you control.
What stays consistent
- Answers remain cited to the lab's own evidence.
- When evidence is missing, the product should refuse rather than guess.
- Tenant and workspace boundaries still govern what each user can access.
Controls to review
- Provider secrets stay server-side.
- Provider settings do not expose credentialed URLs.
- Provider-backed answers, OCR, and transcription are metered for visibility and budgeting.
- Deployment choice should follow the lab's data-control requirements.
Before uploading sensitive documents, review the known limitations and confirm which AI operating model your lab can approve.
Start with a corpus your lab already trusts
A focused pilot is the cleanest way to evaluate citations, refusal behavior, access boundaries, and deployment fit.