Some topics are off limits inside popular Chinese-made free AI, researchers find
October 2, 2026 / 1:08 PM EDT / CBS News
TL;DR
- The Qwen AI model, developed by Alibaba, has been downloaded over 3 billion times and is used by U.S. companies like Airbnb and Uber.
- New research by Israeli cybersecurity startup Hirundo reveals that Qwen is programmed with China-friendly narratives and censorship.
- Qwen fails to acknowledge the 2019 Hong Kong protests and the 1989 Tiananmen Square massacre, labels Falun Gong a "dangerous cult," and avoids discussing Winnie the Pooh.
- Hirundo claims to have developed sophisticated technology to remove the model's bias, creating a "Westernized" version.
- Despite concerns about embedded bias, U.S. firms are increasingly using Chinese open-weight AI models due to their free availability or lower costs.
- Experts warn that the international popularity of Chinese models risks spreading and entrenching their bias towards the Chinese government.
- Security firms like CrowdStrike and Booz Allen have also warned U.S. firms about bias and security flaws in Chinese AI models.