AI Disclosure & Transparency
This pack contains 18 working-draft legal documents. None of them have been reviewed by Texas counsel or counsel licensed in any other jurisdiction. They are a starting point for a real attorney engagement, not a finished publication. Brackets [highlight] indicate custom options to finalize.
SiteFlows uses artificial intelligence ("AI") in several parts of the Service. This notice explains where AI is used, what it does, what data it processes, and what controls apply.
1. Where We Use AI
| FEATURE | WHAT AI DOES | PROVIDER | HUMAN OVERSIGHT |
|---|---|---|---|
| Photo verification | Detects whether a Proof-of-Work photo appears tampered, mis-located, or recycled. Flags suspicious photos for manual review. | OpenAI vision API | Operator decides; AI does not auto-reject. |
| Dispute summary | Generates a neutral summary of the dispute statement, evidence, and proposed resolution. | Anthropic Claude | Summary is suggestive; Operator makes the final call. |
| Pay-rule recommendations | Suggests bonus / penalty rules based on patterns in completed jobs. | Anthropic Claude | Operator must approve before any rule takes effect. |
| Fraud detection | Flags unusual patterns (e.g., GPS spoofing signals, duplicate photos across jobs). | Internal models + Anthropic | Flags trigger review, not action. |
2. What AI Does Not Do
- AI does not classify Workers as employees or contractors.
- AI does not auto-release or auto-deny pay.
- AI does not auto-resolve disputes.
- AI does not set Locked Pay amounts.
- AI does not generate or alter user-submitted photos.
- AI is not used to make any consequential decision about a Worker without human review by the Operator.
3. Data Processed by AI
When AI features are invoked, the following may be sent to the AI provider:
- Photo verification: the photo with its GPS+timestamp watermark. Raw GPS coordinates are not separately transmitted.
- Dispute summary: the text of dispute statements with personally-identifying names redacted before transmission.
- Pay-rule recommendations: aggregated, de-identified pattern data from completed jobs.
- Fraud detection: derived signals (e.g., velocity, duplication scores) rather than raw user data where feasible.
4. Training
Per our agreements with Anthropic and OpenAI, your data is not used to train their models. We do not train any third-party model on Customer Data.
5. Accuracy & Limitations
AI features are probabilistic. They will sometimes flag legitimate photos as suspicious and miss tampered ones. We continuously refine thresholds and review flagged items. The Service does not represent that AI features are perfectly accurate; final decisions remain with the Operator.
6. Right to Human Review
If an AI suggestion affected a decision about you (e.g., a flagged photo led to a dispute), you may request a human review by contacting disputes@siteflows.ai. Human review is provided free of charge.
7. Synthetic Content
For our policy on AI-generated or AI-modified content within the Service, see the Synthetic Content Statement.
8. Regulatory Posture
SiteFlows monitors and seeks to comply with applicable AI regulation, including the Texas Responsible AI Governance Act (TRAIGA), the Colorado AI Act, California SB 942, the Utah AI Policy Act, and (where applicable) the EU AI Act. [Counsel: confirm consequential-decision determinations and any required notices.]
9. Contact
Email privacy@siteflows.ai for AI-related questions or to request a human review.