Three hikers had to be rescued from California’s Mount Shasta this week after leaning on Google’s Gemini chatbot to plan their expedition, according to TechCrunch AI, which cited a Chicago Tribune report on the incident. The Siskiyou County sheriff’s office says the AI advised the group to pack far less food and water than they actually needed. When an 8-hour climb turned into a multiday ordeal, that shortfall became a real problem.
Here’s what happened, per the account TechCrunch AI relayed.
📍 What went wrong
- The three young men started their climb at 3am.
- Standard guidance tells Shasta hikers to turn around if they haven’t summited by noon. This group didn’t reach the top until 7pm.
- They tried to descend in the dark, then called the sheriff’s office asking for directions.
- They spent the night in Mud Creek Canyon before Forest Service rangers and volunteers reached them the next morning.
Nobody died, and that’s the good news. But the sheriff’s office pointed directly at the planning phase, saying the hikers “were advised by Gemini to bring far less food and water than their group required, especially when their planned 8-hour ascent became a multiday ordeal.”
🧭 Why this matters
Gemini didn’t push anyone up the mountain. The decision to keep climbing seven hours past the recommended turnaround time was human. But the food and water miscalculation started at a screen, and that’s the part worth sitting with.
General-purpose chatbots sound confident whether they’re right or wrong. They don’t know your fitness level, the current snow conditions, the weather window, or how your specific group moves on steep terrain. They generate a plausible answer, not a vetted one. For a restaurant recommendation, plausible is fine. For a 14,000-foot volcano, plausible can put people in a canyon overnight.
What stands out here is the mismatch between how these tools present information and how much authority people hand them. A confident paragraph reads like expertise. It often isn’t. The gap between those two things is exactly where this trip went sideways.
⚠️ The pattern to watch
This is part of a bigger trend. As AI assistants get woven into travel apps, mapping tools, and search, people increasingly treat a generated itinerary as a finished plan rather than a rough first draft. The risk isn’t that the AI lies on purpose. It’s that it fills gaps with reasonable-sounding guesses and never flags what it doesn’t know.
It is always advisable to call the local USFS Mount Shasta ranger station ahead of your trip to ensure you have the most accurate information, and to never rely solely on AI for your trip planning.
That advice scales well beyond hiking. Any time an AI gives you a plan where the cost of being wrong is high, treat it as a starting point and verify with a real authority. A ranger station. A doctor. A licensed professional. The chatbot is the intern who read everything and understood some of it, not the expert signing off on the decision.
🔦 What to expect next
Expect more stories like this before the guardrails catch up. As Gemini and its rivals get pushed deeper into everyday planning, the failures will show up in higher-stakes settings than picking a lunch spot. This is significant because it’s an early, concrete example of AI advice creating physical risk, not just a bad answer on a screen.
For now, the practical takeaway is simple. Use these tools to draft, brainstorm, and speed up the boring parts. Then check the pieces that could hurt you against a source that actually knows the terrain. The three hikers made it home. The lesson they paid for is one the rest of us can get for free.
More details are available in the original TechCrunch AI report.