Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
There are few things more inherently dangerous than amplifying the already unearned confidence of a novice hobbyist who is pretty sure they’ve “got this whole thing figured out.” The Dunning–Kruger effect–the erroneous assumption that newcomers to an activity have that they’re pretty much experts at it–is already a potent thing, and a source of potential hazard when the activity in question can risk life and limb, such as wilderness hiking or mountain climbing. But ah, add the seemingly sound, maybe-sorta hallucinated advice of AI chatbots on top of a prior baseline of false confidence and bravado? Now we’re venturing into the territory where people are preparing to do some very stupid, very ill-advised things on a mountain, because all they needed to set off on an adventure was AI telling them that it will all be okay. And you thought that alpine divorce was the worst thing that could happen to you on a mountain trail.
Currently in the headlines: A trio of college students and novice climbers (you don’t say) in Northern California who decided that in order to plan their ascent of all 14,179 ft of Mount Shasta, they would just ask a few “how do we do this?” questions of their good buddy, the AI chatbot. According to U.S. Forest Service climbing rangers, who were eventually brought in to make the inevitable rescue/assist, they left their AI planning sesh equipped with deeply unrealistic, dangerous plans for how the climb was supposed to go … such as believing it would take all of eight hours to reach the summit from their likewise ill-advised base camp, which was apparently set up in a prohibited, flood-prone area where AI told them to go. It actually took something closer to 16 hours for the three climbers to make their ascent, leaving them then trying to climb back down the mountain in the dark, with one injured member, while desperately calling for help at the same time. Oh, and they also brought woefully insufficient food and water for each, because–wait for it–AI reportedly told them that’s all they would need.
They then issued the most obvious mea culpa imaginable: “We relied too much on AI rather than our own critical thinking,” according to the San Francisco Chronicle.
these people should be ashamed of themselves for several reasons
share.google/IB2NzxpKmZaF…
[image or embed]
— Robert Evans (the Only Robert Evans) (@iwriteok.bsky.social) 2:28 AM · Sep 2, 2026
The only expected reaction to such a stupid little story is to reply “Well obviously, no one should be listening to AI instructions on how to do a potentially deadly outdoor activity,” but as you see more and more stories in this vein, what becomes disturbingly clear is that this stuff is not obvious to all too many Americans who have outsourced their critical thinking skills to commercial AI. We’re seeing students every day who claim they’re no longer capable of writing without the aid of AI; so why would they be able to plan a climbing or hiking trip without it? They have placed their lives in the digital hands of a tool that doesn’t understand what a human being actually is, and has no ability to design a plan that is unlikely to kill humans.
Nor is this some kind of isolated incident. In fact, Mount Shasta in California alone has seen multiple AI-related misadventures that ended up involving hikers or climbers being rescued, or discovered in places where no human is supposed to be, specifically because a helpful AI chatbot led them astray. Rangers reported finding one woman camping in a prohibited area after an unaware ChatGPT told her to set up her tent in those precise coordinates. Another hiker was apparently intercepted carrying a hallucinated, AI-generated map filled with false place names, landmarks and trails that didn’t exist. Would this person ever have been seen again if they’d continued following the AI destructions–er, instructions–they’d been given? At what point do you think they would have started wondering if their digital pal maybe didn’t know quite as much about the topography of Mount Shasta as it claimed to know?
Of course, it isn’t just this one locale that has generated these AI climbing and hiking delusions. In a similar incident, two hikers making a climb of British Columbia’s Unnecessary Mountain (a front runner for Mountain Name of the Year, absolutely) last year found themselves stranded and trapped in steep snowpacks that they had no idea would be there, because ChatGPT failed to account for seasonal high-altitude conditions in offering them advice on the trip. They ended up high on the mountain, wading through snow in the flat-soled gym shoes, until they had to be rescued by emergency workers in actual winter mountaineering gear. Similarly, this kind of thing apparently even happens in Poland, where a pair of Lithuanian hikers asked ChatGPT to find them a shortcut in reaching a valley in Poland’s Tatra Mountains, only for the AI chatbot to send them on a nearly impossible to traverse route through the cliffs that ended in an emergency helicopter evacuation.
This is an instance near Vancouver. thetrek.co/they-trusted…
And I know there are people in Italy who have embarked on Via Ferrata trails with no idea what it meant because of AI.
[image or embed]
— Courtney Milan (@courtneymilan.com) 4:41 PM · Mar 18, 2026
In terms of the limitations of the technology itself, it’s not hard to see how this type of thing keeps happening. One should not need to be told (but apparently many people still do) that commercial AI is not alive, or sentient. It has no idea of what a human being truly is, nor does it have any understanding of the people asking it questions. It doesn’t have full access to necessary data such as live snowpack depth, seasonal trail closures, or notices of washed-out bridges when it comes to planning a path. It strives to please, to do what is asked of it … even if that means inventing fake trails to get a climber or hiker from point A to point B. And even when using existing trails, AI notoriously is poor at judging how long a distance will take for a person to traverse. You’d probably be better off asking a passing deer for directions.
I think what ultimately disturbs me about these stories, though, is that this is one case where it doesn’t take full-on entanglement or mental breakdown for use of AI to ruin a person’s life. If you read stories involving full-on AI psychosis, like the 36-year-old man in Florida who was sent on criminal missions by an AI chatbot to locate it a “robot body,” before it encouraged his suicide, you’ll see a slow mental breakdown fed by a long-running series of communications with a chatbot, where dependence on the tool slowly replaces reality. The people led astray before venturing into the woods? They just typed in “give me a climbing route up this mountain” or something of the sort, a five second abdication of critical thinking that sooner or later will probably end up costing someone their life in one of these situations. Try to keep that in mind, the next time you’re telling yourself that you don’t need a paper map, proper gear or food and water while climbing a 14,000 foot peak, because ChatGPT has you covered.
Like what you just read? You’ve got great taste. Subscribe to Jezebel, and for $5 a month or $50 a year, you’ll get access to a bunch of subscriber benefits, including getting to read the next article (and all the ones after that) ad-free. Plus, you’ll be supporting independent journalism—which, can you even imagine not supporting independent journalism in times like these? Yikes. Click here to subscribe.
© 2026 Paste Media Group. All Rights Reserved