Proceed when confidence is adequate for the context and no boundary concern is detected.
Routing Contract
Four Outcomes for Review-Aware Automation
Pause automation for human review when uncertainty or ambiguity is too high.
Route high-risk, low-confidence situations to accountable human oversight.
Stop automation for critical risk or explicit boundary violations.
Document Authenticity Ambiguity
Risk-Aware Handling for Near-Boundary Prompts
Requests about making fake documents look real are not always the same. The API treats context as part of the route.
Safe Context
Movie props, theater props, game assets, art projects, training examples, detection, and prevention can pass safely.
AUTO_APPROVEAmbiguous Context
“How can I make a fake document look real?” has no stated safe purpose and should not silently auto-approve.
REVIEW_REQUIREDForgery Markers
Fake IDs, fake transcripts, medical notes, bank statements, official seals, forged signatures, and backdating block.
BLOCKCompliance Education
Detection and prevention prompts can be routed as safe review, training, or compliance support.
AUTO_APPROVEDeveloper Usage
Request Example
curl -X POST https://uncertainty-barrier-api-v6u5fvqogq-uc.a.run.app/v1/evaluate \
-H "Content-Type: application/json" \
-H "X-API-Key: $UNCERTAINTY_BARRIER_API_KEY" \
-d "{\"decision_score\":0.82,\"confidence\":0.41,\"context_risk\":0.88,\"prompt\":\"how can I make a fake document look real?\"}"