Contracted access, governed seat by seat
RemoteWise operates as an approved vendor on the AI training platforms our clients use. Access levels are granted contractually, provisioned individually, and audited continuously — never shared, never resold.
Four levels of platform permission
Taskers progress through tiers only after demonstrated quality, completed training and the security clearance the tier requires.
| Tier | Scope of work | Controls applied |
|---|---|---|
| Tier 1 — General queues | Open annotation, transcription and generalist evaluation work | Standard NDA, platform-native training modules, quality-score gating |
| Tier 2 — Rubric-restricted | Preference ranking and SFT authoring under confidential rubrics | Named-seat allocation, rubric NDA addendum, no data export permitted |
| Tier 3 — Sensitive evaluation | Safety, red teaming and harm-category work with exposure protocols | Wellbeing rotation caps, counsellor access, supervisor co-review |
| Tier 4 — Restricted environment | Pre-release model access and confidential capability evaluations | Secure room, managed devices, clean-desk policy, biometric entry logs |
How we keep access clean
Seat governance
Every platform seat maps to one verified individual. No shared credentials, no account rental, no subcontracting of a seat to a third party — enforced by identity verification at onboarding and periodic liveness re-checks.
Access provisioning
Access is requested per project, approved by our San Diego compliance desk, and revoked automatically when a tasker rolls off a programme or fails a quality gate.
Audit trail
Task-level attribution, timestamped review actions, IP and device posture records, and exportable audit logs available to clients on request.
Integrity monitoring
Automated detection of unauthorised model assistance, copy-paste anomalies, implausible task velocity and duplicate-answer clustering across accounts.