AnnouncementTurnkeeper is working toward an open standard for sharing safety intelligence across platforms.

Turnkeeper Ward

Safety Intelligence workspace

Contract workspace for design partners: scenarios, inspector, and research packs on local fixtures only. Not GA.

Back to story mode

Synthetic walkthrough — opaque IDs only; no raw content. Findings never authorize automatic enforcement. Not GA.

Current scenarioMulti-detector: Sentinel + Hive CSE–shaped + ToxMod-shaped

Synthetic case loaded. · Pipeline: ready for customer revalidation

  • OSS text detector
  • CSE-shaped classifier
  • Voice session risk

1. Observe

Hypothesis · not a verdict · not enforced. Bounded events the platform chose to send.

  1. R-01 · boundary testing · OSS text detectorMay 12, 2026, 9:12:00 PM UTCcand_multi_sentinel_bnd01 · Payload closed
  2. R-02 · boundary testing · CSE-shaped classifierJul 4, 2026, 4:03:00 PM UTCcand_multi_hive_groom01 · Payload closed
  3. R-03 · off platform move · Voice session riskAug 7, 2026, 8:41:00 AM UTCcand_multi_toxmod_off01 · Payload closed

2. Connect

Candidate trajectory across time — not a verdict.

  1. boundary testingobserved
  2. secrecy or isolationobserved

3. Review

Potential grooming trajectory

Potential grooming trajectory. Requires specialist review. Trajectory includes 2 observed stage(s).

Confidence
low
Severity
elevated
Enforcement authorized
false
Recommended workflow
specialist review

Review resolved. Outcome does not execute application actions.

Recorded: substantiated. Signals from text affinity, CSE text classification, and voice session detectors reconstruct one longitudinal hypothesis. Specialist review required. No automatic enforcement.

4. Authorize

Customer action

Disabled until customer revalidation. Review approval alone never executes an application outcome.

Recorded action: none. No customer application action recorded. Review approval does not execute actions.

Same contracts as the hosted Ward design-partner path.

Hosted ingest and cases live under Ward safety intelligence APIs on allowlisted projects. This page exercises the contracts with local fixtures only—no automatic enforcement and not general availability.