AI Got It Wrong
Short answer: When AI answers a question about a rule, it gives the answer that is usually true somewhere. Organizations need a short written list of questions AI never settles alone (policy, pay, legal terms, safety, customer commitments), each with a named source of record and owner, checked before anyone acts.
What happened on Mount Shasta?
Rangers on Mount Shasta in Northern California are meeting a new kind of visitor: one who planned with a chatbot and believes it. A woman camped beside a “No Camping” sign and insisted AI had told her she could. A man arrived at a trailhead with an AI-generated map that put landmarks on the wrong side of a nearby mountain. Three young hikers needed a rescue after, according to the county sheriff, the chatbot they planned with told them to bring far less food and water than the climb required.
Siskiyou County Sheriff Jeremiah Larue drew the line in one sentence: “AI can be a good tool, but it shouldn’t be used as the expert.”
Why is this an organizational problem, not a personal one?
Each visitor made a reasonable-looking move. They asked a capable tool a direct question and acted on a fluent answer. Nobody had told them which answers needed a second source, and the tool gives no signal that separates general knowledge from a local rule.
Employees are in the same position every day. Whether an expense is allowed, what a contract permits, which customer data may be shared, what was promised to a client: AI will answer all of these in the same assured tone. Its answer reflects what is common, not what your organization decided. If the organization has not said which questions need the source of record, each person decides alone, and some will decide the way the camper did.
What should leaders put in place?
Write the list. Name the categories where AI never gets the last word: policy, pay and benefits, legal terms, safety, and commitments made to customers. Keep it short enough to remember.
Name the source and the owner. For each category, point to the document of record and the person who maintains it. An answer is checked against that source, not against a second chatbot.
Build the check into review. When AI-assisted work touches a listed category, the reviewer confirms it against the source before it ships. The check becomes part of the process instead of something each person has to remember.
The goal is not less AI. A ranger would happily let a hiker sketch the trip with a chatbot. The sign at the campsite still decides where the tent goes.
Most teams have not yet mapped where they rely on AI for answers it cannot know. See where your team stands →
Keep reading: When nobody can say what AI touched · Use AI to find the expert, not to replace one
This is the thinking behind a Trusted AI Culture.
Sources: High Country News, "AI gets people into trouble in the backcountry," October 23, 2026. KMAX/KOVR via CNN Newsource, "Hikers who used AI to plan mountain trip had to be rescued, authorities say," September 3, 2026.
Written with AI in the loop: my idea, AI drafted and sharpened, my judgment on the way out. Every word is mine to stand behind. — Darren