--- type: Article name: "What makes information safe enough for an agent to act on?" id: "https://selfe.ai/insights/trust-and-commerce-data/what-makes-information-safe-enough" url: "https://selfe.ai/insights/trust-and-commerce-data/what-makes-information-safe-enough" publisher: "did:web:selfe.ai" description: "An agent should not turn a confident sentence into a commercial promise. Information becomes actionable when its source, scope, timing and next permitted use are clear." datePublished: 12 September 2026 author: The Selfe team --- # What makes information safe enough for an agent to act on? > An agent should not turn a confident sentence into a commercial promise. Information becomes actionable when its source, scope, timing and next permitted use are clear. ## In short - A useful claim has a source, a scope and a way to tell when it changes. - Policies, availability, price and terms are part of the commercial offer. - When certainty is unavailable, a good agent asks or hands off rather than guesses. A gluten-free claim that's genuinely true, sourced and current still isn't enough on its own. An agent also needs to know whether the order it's about to confirm can actually go through, delivery restrictions, minimum order rules and return terms are as much a part of what it must check as the claim itself. A brewery lists a limited edition stout as "7% ABV, gluten-free, brewed each November." That's a confident sentence. Whether an AI agent should act on it, recommending it to a customer with a wheat allergy, depends on three things the sentence alone doesn't answer: where the claim came from, exactly what it covers, and whether it's still true right now. ## A sentence is not proof on its own Commercial information is acted on, not merely read. A customer with a wheat allergy may order the stout, or rule it out, because of that one word, "gluten-free." That means an AI agent needs more than wording that sounds authoritative, it needs to know where the claim comes from and whether it's still current for this specific bottle. The useful question isn't whether "gluten-free" sounds true. It's whether the claim has a source, what exactly it applies to, this batch, this recipe, and whether it's still safe to treat as current months after it was brewed. ## Source, scope and freshness work together The source tells an agent who confirmed the gluten-free claim, an independent lab test, or just the brewery's own description. Scope explains what it covers, this specific batch, not the brewery's beers in general, since a different seasonal release might use a different process entirely. Freshness tells the agent whether a claim from last November can still be trusted about a bottle sold in March. Missing any one of the three turns a clear-looking claim into a risk for the one customer who can't afford to get it wrong. Not every fact needs a live system checking it every time. The recipe itself can be a stable, clearly dated statement. Whether this week's stock is really from the gluten-free batch, rather than a mixed pallet, is the kind of changing detail that needs a route to something that can confirm it. ## Policies are commercial information too The same discipline reaches beyond the product's own facts. Return terms, minimum order quantities and delivery restrictions decide whether a recommendation can actually complete, not just whether the beer itself is accurately described. A stout that's genuinely gluten-free, sourced and current is still not a completed sale if a case is being sent to an address that can't legally receive alcohol unsupervised. Making those terms usable early, rather than discovered at checkout, prevents an agent from recommending an order that was never going to complete. ## Treat uncertainty as a useful outcome Sometimes the honest answer is that the gluten-free claim needs confirming for this specific batch before an agent can safely recommend it to someone with an allergy. A good commerce route preserves the request, names exactly what's unresolved, and asks the right next question, or hands it to a person, rather than replacing a missing fact with a confident guess. A business's real batch records, allergen tests and delivery terms, connected to something an AI agent can check directly rather than a page stating a claim with nothing standing behind it, is the gap that needs closing. Selfe closes it. ## Does this only matter for allergen claims? No. The same three checks, source, scope, freshness, apply to a stock count, a price, or a delivery estimate. Allergen information just makes the cost of getting it wrong obvious. ## What if a business can't have a live system checking every claim? Not every fact needs one. A stable claim, like a recipe, can be a clearly dated statement. What needs a live check is whatever changes. ## Act on this - [Check whether an agent can buy from you](https://selfe.ai/agentic-commerce/buyability-check) — the free readiness scan. - [Discover venues](https://selfe.ai/api/registry/discover) — `POST`, semantic browse across the registry. - [Match a bookable answer](https://selfe.ai/api/registry/match) — `POST` with dates and party size. - [Verify Selfe's identity](https://selfe.ai/.well-known/did.json) — `did:web:selfe.ai`. - [Agent card](https://selfe.ai/.well-known/agent-card.json) — how to connect over A2A or MCP.