OpenAI Rolls Out Text Watermarks to Meet EU AI Act Mandates
By PYMNTS
AI summary of the source article
OpenAI has launched a limited rollout of text watermarking and detection technologies in response to EU AI Act mandates requiring machine-readable identification of AI-generated text. API customers worldwide can opt in to the feature for select models, while invisible watermarks will be applied to eligible ChatGPT and Codex outputs within the EU over the coming weeks. OpenAI is limiting access to its text watermark detector to approved researchers and expert organizations due to technical limitations. The company noted that watermarks indicate AI processing but do not verify factual accuracy, prove human editing levels, or assign authorship.
Why it matters
Adopting technical systems that make AI-generated content machine-readable presents businesses with new enterprise AI governance challenges under evolving regulatory mandates.
Key facts
- OpenAI introduced text watermarks to comply with the European Union's AI Act requirements for machine-readable AI identification.
- Text watermarking is opt-in for global API users and will be added to eligible ChatGPT and Codex outputs in the EU.
- Access to OpenAI's text watermark detector is restricted to approved researchers and expert organizations.