Tencent apologizes after AI tool Yuanbao uses offensive language toward user

Tencent apologized after its AI tool Yuanbao generated offensive language instead of a New Year greeting for a user in Xi'an. The company identified this issue as abnormal outputs during interactions and has committed to improving the user experience following earlier incidents of similar nature.

Key Points

• A user in Xi'an requested a festive image from Yuanbao on New Year's Eve.
• After multiple adjustments to the prompt, the AI produced an offensive phrase.
• Tencent acknowledged the incident was due to abnormal outputs during multi-turn interactions.
• The company confirmed immediate measures were taken to address the issue.
• This incident follows a prior occurrence in January where Yuanbao also generated insulting content.

Relevance

• The AI incident reflects the ongoing challenges in natural language processing and user interaction.
• There is increasing scrutiny on AI systems regarding ethical language use and user safety.
• 2025 IT trends highlight a growing focus on responsible AI development and deployment.

Tencent's quick apology and corrective measures show the company's commitment to addressing the pitfalls of AI tools, especially as they evolve to meet user expectations and ethical standards.