OpenAI Implements ChatGPT Watermarking in EU

OpenAI has officially launched a default watermarking mechanism for ChatGPT outputs, specifically mandating compliance within the European Union under current regulatory frameworks.

The top 3

  1. EU AI Act Compliance Requirements: The watermarking initiative serves as a direct response to the EU AI Act, which necessitates transparency labels for synthetic content to mitigate misinformation risks.
  2. Technical Limitations of Current Watermarking: OpenAI acknowledges that these watermarks are not foolproof and can be easily circumvented through paraphrasing, image editing, or simple re-prompting techniques.
  3. Exclusion of Non-EU Markets: The default watermark policy is currently restricted to EU territories, leaving users in North America and Asia without the same mandatory provenance labeling.

Sources

Open the full topic