# LLMs respond differently to harmful prompts when AI watermarking is used

> 标题为英语

> 全文为英语

**Published:** 2026年9月17日
**Author:** Dan Goodin
**Source:** Ars Technica — AI
**Language:** en-US
**HTML:** https://aiqazaqstan.com/zh/articles/llms-respond-differently-to-harmful-prompts-when-ai-watermarking-is-used

In response to a new European Union law, AI platforms are implementing new schemes for watermarking the content they generate. Anthropic recently disclosed its future Claude models will use SynthID-Text, an approach Google created and released as open source. It uses a secret…

> 内容来自原始来源： **Ars Technica — AI**
> [在 Ars Technica — AI 阅读全文](https://arstechnica.com/security/2026/09/ai-text-watermarking-can-make-models-more-vulnerable-to-adversarial-prompts/)

**标签：** #Anthropic

---

*AI Qazaqstan — non-commercial informational portal about AI in Kazakhstan. Machine translations and summaries are AI-generated.*