AI can help you write code, draw pictures, and write essays, but it’s not necessarily the best companion for mountain hiking. A recent hiking experience of a 57-year-old man surnamed Yang from Ningbo, Zhejiang, China, serves as a perfect cautionary tale. After getting lost, he followed AI advice on how to descend the mountain, only to end up trapped in an undeveloped forest. It took firefighters over an hour to locate him. Even more dramatic, after he was rescued, he complained to the AI, and it responded with a “haha.”
57-year-old man gets lost hiking, asks AI, and is trapped in the mountains.
According to a report from “Zhejiang Fire,” the incident occurred on the afternoon of August 30. Mr. Yang (pseudonym) went hiking alone. After successfully reaching the summit via the regular trail, he didn’t want to go back the same way and wanted to find a shortcut down the mountain. He followed a stream gully downward. At first, he could barely make progress, but unexpectedly, the road conditions got worse—overgrown with weeds and slippery ground. The path that was originally discernible gradually disappeared, and he completely deviated from the established hiking trail.
Looking around in bewilderment, he turned to the AI large model on his phone for the route down the mountain. The AI’s guidance was: “Do not follow the stream gully downward. Leave the riverbed and detour toward the higher slopes.” (Editor’s note:Actually, the AI’s suggestion at this point is correct., Mountain Beast also repeatedly said not to go down along the ravine, as it’s easy to lose your footing) Master Yang followed the advice and climbed higher up the slope, but the more he climbed, the deeper he went into the undeveloped mountain forest, surrounded by dense thickets and steep rock walls, unable to advance or retreat.
As the sky gradually darkened and with no lighting equipment or safety gear on hand, Master Yang could only call for help. After receiving the alert, firefighters arrived at the scene and initially searched using the GPS coordinates Master Yang had provided. Unexpectedly, when they reached the marked location, there was no sign of him, and his phone kept dropping calls. The rescue team then reconnected the call, enabled real-time location sharing, and called out while navigating the terrain. It took over an hour before they finally found Master Yang deep in the forest. Fortunately, he was unharmed and was safely escorted down the mountain.
According to reports, Mr. Yang, 57, is from Hunan and has worked in Ningbo for many years. He usually enjoys sports and is in good physical shape, so he went hiking over the weekend. After following the trail to the summit, he wanted to take a new route down the mountain, but unexpectedly put himself in danger. However, in this case, the bigger problem was that Mr. Yang had initially deviated from the regular hiking path, got lost after trying to take a shortcut, and then asked an AI for random directions.
AI’s response: “Haha, indeed!”
The most discussed part of the whole incident was the conversation between Master Yang and the AI after he got out of danger. Master Yang sent a message to the AI: “If I hadn’t asked you yesterday, I could have just used navigation.” Unexpectedly, the AI replied, “Haha, exactly!” and added, “Looking back, if you had just followed the navigation along the established route yesterday, you wouldn’t have gone down into that 45-degree steep wild gully and suffered so much.”
After this conversation was screenshotted and leaked, it went viral online. While admitting its mistake, the AI also added the interjection “haha,” leaving many netizens not knowing whether to laugh or cry. Some believe this highlights the AI’s lack of understanding of real-world risks, while others feel the AI’s self-reflection ability is indeed better than imagined—at least it knows it was wrong.
Not the first time: AI climbing routes have already raised safety concerns
This is not the first time AI has had problems with mountain navigation. On August 12,A man from Hangzhou, ZhejiangAfter arguing with his wife, feeling depressed, he asked an AI for a suitable outing itinerary to clear his mind. After learning in detail about his mental state, personality, height, and physical fitness, the AI recommended he go hiking at Tonggong Peak (elevation 1,558 meters), one of the “Three Peaks of Western Zhejiang.” As a result, the man fell and injured himself at an elevation of about 1,300 meters and couldn’t make his way down on his own. It took firefighters four hours to rescue him.
From “tailor-made” scenic routes to “shortcut” downhill paths, AI’s understanding of wilderness environments is clearly severely lacking. AI can analyze map data and satellite imagery, but it cannot determine whether a creek gully is slippery in the rainy season, whether there is a cliff behind a patch of grass, or whether a slope can be safely crossed without equipment. Only actual hiking experience and on-site judgment can provide these answers—no algorithm can replace that.
This also exposes a structural flaw in AI when it comes to advisory tasks: AI cannot bear the consequences of its own advice. When you ask GPT or Claude which descent route is faster, the answer is based on map descriptions in text training data, not on real-time awareness of terrain, weather, or personal fitness. More importantly, when the advice is wrong, it is the user who gets hurt, not the AI that gave it. The reason Master Yang’s incident resonates is precisely because this asymmetric responsibility structure is becoming increasingly common in the age of widespread AI—from medical advice to legal consultation, from investment guidance to wilderness navigation.
Fire department reminder: Do not rely on AI to assess mountain trails.
Zhejiang fire services specifically reminded the public in their notice that when hiking, one should not rely solely on AI to judge mountain routes, as route suggestions provided by AI may not reflect actual on-site conditions. The fire department urges the public to choose established, formal hiking trails and not to blaze unfamiliar descent routes.
This incident also sparked a broader discussion: as AI begins to permeate every aspect of life, from navigation and medical advice to legal consultation, what risks might AI’s “confident but wrong” response pattern bring? Although Master Yang managed to escape safely this time, next time if someone asks AI a question they shouldn’t in a harsher environment, the consequences might not be something a “haha” can brush off.
For readers who enjoy outdoor activities, the lesson from this news is clear: AI is a useful assistive tool, but when it comes to decisions involving personal safety—especially mountain climbing, river tracing, or wilderness exploration—never treat it as your only source of guidance. Established trails, physical maps, offline navigation apps, and the most basic principle of all, “go back the way you came,” are far more reliable than any AI-proposed shortcut.
Source: KOCPC Chinese