Academy
6. Limits of a Green Badge
Recognise that a green badge means consistency with visible sources, request, and tools—not absolute truth—and that human judgment remains necessary for decisions and commitments.
A green badge means the answer is consistent with what TrustLayer can see — your sources, your request, and the tools the AI used. It does not mean the statement is true in the real world. A claim can be perfectly supported by the documents you provided and still be factually wrong if those documents are outdated, incomplete, or themselves incorrect. Green lowers the risk; it does not eliminate it.
TrustLayer runs five independent controls: source anchoring (are claims found in the provided sources?), prompt fidelity (does the answer cover your request and respect constraints?), action claims (did claimed actions actually happen?), code guard (dangerous code patterns), and repetition (recycled content). A green score means all five passed — but each control only checks what it can observe. None of them fact-checks against the entire world's knowledge.
Consider a concrete scenario: you upload a product sheet from 2023 and ask the AI for the current price. The AI quotes the price from that sheet. Source anchoring passes — the claim matches the document. Prompt fidelity passes — the question was answered. But if the price changed in 2024, the answer is wrong despite a green badge. The badge confirmed consistency with visible sources, not real-world accuracy. This is exactly why your judgment on decisions and commitments remains necessary.
Before acting on any green-badge answer that will drive a decision or commitment — a quote to a client, a legal clause, a published figure — cross-check the underlying sources yourself. Click the badge to see which claims were anchored and to which documents, then verify those documents are still current.
Grounding your questions with up-to-date documents or web results raises the value of a green badge, because source anchoring has more material to verify against. An ungrounded answer — one with no attached sources — cannot be anchored at all, so the badge has less to work with. When accuracy matters, always attach relevant, current documents or enable web search before asking.
An orange or red badge is an explicit invitation to check. But a green badge is not an invitation to stop thinking. Treat green as "lower risk, proceed with normal professional diligence" — not as "guaranteed correct."
Open a chat in OUPI, attach a document you know well, and ask a question whose answer you can verify independently. Click the green badge to inspect each control's result. Then ask yourself: does the answer match reality, or only the document? This exercise builds the reflex of reading the badge critically.
A green badge confirms internal consistency — the answer aligns with your sources, respects your request, and matches real tool actions. It does not certify real-world truth. Sources can be outdated, incomplete, or wrong. Always ground your questions with current documents to strengthen the badge's value, click the badge to inspect the details, and apply your own judgment before any decision or commitment. Green means "verified against what's visible" — the rest is up to you.