OpenAI is set to implement invisible textGrain watermarks in the outputs of ChatGPT for users in the European Union. This move aims to enhance the traceability of AI-generated content, addressing concerns over misinformation and content authenticity.
Interestingly, unlike its competitor Anthropic, OpenAI will allow API users outside the EU to opt out of this watermarking feature. Preliminary tests indicate that the watermarking can achieve detection rates as high as 95 percent, although this effectiveness diminishes significantly when a substantial portion of the generated text is altered.
This decision reflects OpenAI’s efforts to comply with regulatory standards in the EU while also considering the preferences of its global user base. As AI technology continues to evolve, such measures are becoming increasingly important in fostering trust and accountability in AI-generated content.
Compiled automatically by the Tech AI Newsdesk from public AI-news sources and summarised in our own words.
