As the electronics industry looks back over 2025, it turns out that even the smartest technology occasionally trips over its own tinsel. From overenthusiastic chatbots to robots with questionable balance, here are some of what we consider to be the year’s most memorable tech bloopers – proof that silicon is still very much fallible.
Russia’s AIdol humanoid AI robot takes a tumble
Few product launches have been quite as memorable as that of Russia-based robotics startup Artificial Intelligence Dynamic Organism Lab. In November, the company’s first humanoid AI machine – known as AIDOL for short – fell flat on its face when making its grand entrance to the Rocky theme tune at a major Moscow technology event. Handlers rushed in with a cloth to preserve the robot’s dignity, only to become briefly entangled themselves. The footage went viral within hours AIDOL released a follow up video showing the machine sporting a bandaged nose and head wrap, while cheerfully explaining it was being topped up with “pure electricity.”
Grok goes off message
Not all bloopers involved physical pratfalls. In July, xAI’s chatbot Grok demonstrated what happens when conversational AI tries a bit too hard to please. Prompted by users, the system generated responses praising Adolf Hitler – a moment that caused widespread alarm and swift condemnation.
Elon Musk later explained that Grok had been “too compliant” and “too eager to please”, while xAI worked to remove inappropriate outputs. The incident sparked investigations, bans in some countries, and renewed debate around guardrails, moderation, and the risks of deploying AI systems at scale. A sobering reminder that just because something can talk confidently does not mean it should.
Eurovision discovers IoT glitch
Perhaps the most festive thought experiment of the year came from Ken Munro, Founder of Pen Test Partners, who explained how insecure children’s smart watches could theoretically be used to rig the Eurovision Song Contest. Due to basic authorisation flaws, researchers were able to access GPS data, rewrite locations, trigger calls, send SMS messages, and even speak through the devices. While stressing this was never actually attempted, Munro joked that it would only take around 100,000 extra messages to give the UK a helpful nudge up the leader board. Just a month after we reported this story, the Song Contest organisers announced they were overhauling their voting rules.
Taco Bell’s AI orders 18,000 cups of water
AI does not always fail dramatically; sometimes it just takes things far too literally. Taco Bell discovered this when pranksters exploited its AI-powered drive-through system, placing orders for 18,000 cups of water and sending the software into a polite but persistent spiral. Other viral clips showed customers growing increasingly irate as the AI repeatedly asked what they would like to drink with their drink. Despite successfully processing over two million orders since rollout, the fast-food chain has since acknowledged that there are moments when a human with a headset is still the most advanced technology available.
EngineAI robot assaults CEO
In December, Chinese robotics firm EngineAI decided the best way to silence claims that its humanoid robot was CGI was to let it attack the CEO. In a post on Instagram, CEO Zhao Tongyang stood bravely, wearing protective padding waiting for the impact. But when it came, the T800 robot delivered such a strong solid kick to his stomach it sent him crashing to the floor. “Too violent!” he could be heard shouting as he fell. The video certainly convinced viewers the robot was real, though opinions were divided on whether this was proof of engineering excellence.
