The built-in AI editor in Telegram turned out to have a 'surprise'. It turned out that under the hood runs the Qwen3 model from Alibaba.

The neural network is trained to follow the ethical norms of China. If you try to write something critical about China, the model simply refuses to work. In the first screenshot — from a week ago, this was noticeable in the responses, and now the AI just gives an error — the second screenshot.

The Qwen model from Alibaba is currently considered one of the most powerful in the world. However, it has a specific feature: it can 'hallucinate' and output recipes for Chinese dishes if it does not understand a complex technical request in another language.


📲 MAX 👮‍♂️ VK ☺️DZEN 🙃 SETKA ✈️ TG #telegram #Qwen3 #AI #ai