Climbers Rescued After Relying on AI for Mount Shasta Trip

Last week, three inexperienced climbers needed help on California’s Mount Shasta after they heavily relied on Google’s Gemini AI chatbot for their expedition, as reported by the U.S. Forest Service. One of the climbers admitted to rangers, “We relied too much on AI rather than our own critical thinking.”

The San Francisco Chronicle detailed how the college students began their journey on Saturday from the Clear Creek Trailhead, located on the southeastern side of the mountain. They intended to camp midway and make a push for the summit early the next day. However, when they encountered difficulties on their way back down, they ended up spending a night outdoors, prompting a response from the Siskiyou County Sheriff’s Office and the Forest Service.

The Clear Creek route is about 8 miles long and ascends 6,500 feet. Once the snow melts, it’s considered one of the easier routes since it doesn’t require special gear like ice axes or crampons. The climbers had prepared for the trip by watching YouTube videos and consulting Google’s Gemini AI for advice on their route as well as the amount of food and water to take. For navigation, they planned to use the AllTrails mobile app.

They successfully reached the summit on Sunday evening, but this was well past the recommended turnaround time from rangers. On their descent, they misinterpreted a suggested route line in AllTrails as an actual trail. Nick Meyers, the lead climbing ranger and director of the Mount Shasta Avalanche Center, highlighted that “those are suggested routes, not trails.”

Unfortunately, things deteriorated from there. The phone they were using for navigation ran out of battery, and a backup power bank failed to connect. By Sunday night, the group was described by the Forest Service as “well off-route,” forcing them to camp outside. They called for help that night, and rangers located them the following morning.

Upon reaching the climbers, rangers found them exhausted and in dire need of food and water. They didn’t have proper clothing or equipment and had set up camp in a restricted area near a freshwater spring, leaving behind trash and scraps. The rangers escorted them back to base camp.

After the rescue, the climbers revealed that the AI had advised them to consume simple carbohydrates instead of fats for their climb, stating that fats “take too long to digest.” This contradicts standard climbing practices regarding nutrition. They were also told they wouldn’t need extra layers of clothing, which left them vulnerable to the cold at night.

Reflecting on their experience, the climbers expressed regret about placing too much trust in the AI. The Sheriff’s Office labeled this dependence as “a critical misstep,” while Meyers bluntly stated, “They were totally blowing it in all regards.”

This incident isn’t isolated; similar issues have arisen this year due to AI-generated trip advice on Mount Shasta. Earlier in the summer, a woman was discovered camping in a clearly designated no-camping area, explaining that “AI told me I can camp here.” In another instance, a man incorrectly used a route printed from ChatGPT that contained numerous inaccuracies. “It was terrible, terrible, not even close to accurate,” Meyers remarked, advising the individual to discard the map.

Meyers noted a concerning trend this season, stating, “This is the first year where we’ve started to have some AI fallout. People are using it for trip planning, asking, ‘Hey, can I do this up here?’ It’s not always correct.”

When people allow AI to take over their decision-making, a phenomenon known as “cognitive offloading,” it can lead to problems — whether it’s inexperienced climbers stranded in the wilderness or young individuals struggling with mental health issues.