TIAKI operates a multi-agent intelligence ecosystem ("AI Outputs") spanning sourcing, provenance, MRV, pricing, sanctions, and Trust ROI scoring. This notice sets out the limits, transparency obligations, and governance applied to those outputs, consistent with the EU Artificial Intelligence Act (Regulation (EU) 2024/1689), in particular Articles 9, 13, 14, 50, and 86.
1. Human-in-the-Loop by Design
AI Outputs are decision-support, not autonomous decisions. Every output with material legal, financial, or safety effect passes through a documented human-in-the-loop adjudication step, including hash-divergence reconciliation, Trusted-Flagger-style escalation, and reviewer override.
2. "Architected to Meet" — Phrasing Rule
TIAKI uses the construction "architected to meet" or "designed to meet" criteria. This phrasing is deliberate: it conveys engineered alignment with a regulatory, technical, or contractual standard. It does notassert certification, accreditation, attestation, or guarantee of outcome.
3. Not Financial, Legal or Regulatory Advice
AI Outputs — including Trust ROI scores, capital-cost compression bands, sanctions contagion graphs, MRV mass-balance results, and pricing ticks — are illustrative synthetic or reference outputs unless otherwise marked. They are notinvestment, legal, tax, regulatory, actuarial, or audit advice and must not be relied upon as such.
4. Provenance, Data Mix & Watermarking
Every AI Output is accompanied by a provenance footer disclosing the data mix (public / synthetic / confidential), regulatory anchors, and a Source Authenticity Certificate. Pre-validation outputs carry a "Pending domain validation" watermark. Generated synthetic content is identifiable as such, consistent with EU AI Act Art. 50(2).
5. Known Limits
- Models may produce plausible but incorrect content; outputs require expert review.
- Sovereign source registers are extensive but not exhaustive of every primary regulation, gazette, or sanctions list.
- Reference scoring uses deterministic seeds where live calibration is pending.
- Outputs are time-sensitive and may not reflect intra-day regulatory or market events.
6. Prohibited Uses
You must not use AI Outputs to take fully automated decisions producing legal or similarly significant effects on natural persons, nor to derogate from the human-oversight controls TIAKI has architected. Training third-party models on AI Outputs is prohibited.
7. Incidents & Feedback
To report a suspected error, drift, hallucination, or governance concern, contact investors@tiaki.ai. Material model incidents are logged in TIAKI's AI Governance register consistent with EU AI Act Art. 12 and 17 record-keeping requirements.