OpenAI built the technologies in response to the European Union’s AI Act, which requires generative artificial intelligence providers to make AI-generated text identifiable in a machine-readable way, according to the article.
OpenAI is enabling API customers globally to opt in to text watermarking for select models, starting today, with text watermarking remaining off by default in the API, the article said.
Over the coming weeks, OpenAI will add an invisible watermark to eligible ChatGPT and Codex text output in the EU, per the article.
We’d love to be your preferred source for news.
Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!
In addition, OpenAI is opening applications to access its text watermark detector. The company is limiting access to this tool to approved researchers and expert organizations that can help OpenAI evaluate and improve the technology, according to the article.
“Text watermarking and detection remain early technologies with significant limitations, and views about their benefits and responsible uses are still developing,” OpenAI said in the article. “Our phased approach reflects both the EU AI Act requirements as well as the technology’s limitations, with an emphasis on transparency about what a text watermark can and cannot tell people.”
OpenAI said in the article that it is important to understand that a watermark can indicate that an OpenAI system generated or processed part of a passage, but not how much human editing went into it; a watermark does not establish ownership or responsibility; a watermark does not associate a user with the text; and a watermark does not verify the accuracy of the passage. In addition, the absence of a watermark does not prove the passage was authored by a human, the article said.
“We’ll continue improving detection, studying how watermarks withstand editing and translation, and exploring ways to distinguish AI assistance from AI authorship more meaningfully,” OpenAI said in the article.
The limited rollout announced Monday applies only to OpenAI’s text provenance tools. They do not affect the company’s verification tools for audio and images, which are publicly available, per the article.
Anthropic announced in August that it had begun labeling its models’ content as AI-generated to comply with new EU regulations, PYMNTS reported. The company said that its Claude models launched in the EU on or after Aug. 2 would support machine-readable marking at launch, with watermarks embedded into AI-generated text.
PYMNTS reported in August that the adoption rules requiring technical systems that make AI-generated content detectable present businesses with a new enterprise AI governance challenge.
Add as Preferred Source
