OpenAI / ChatGPT

We're expanding our approach to content provenance to include text in response to EU regulatory requirements, while recognizing the significant limitations of current text watermarking technology. Our tools already help verify whether an image or audio file was created with our models. This work builds on those efforts to help people better understand when content may have been generated or edited with an OpenAI model. In the EU, we’ll start watermarking eligible text from ChatGPT and Codex over the coming weeks to comply with the EU AI Act. Customers using our API can turn on text watermarking for select models worldwide today.

We're expanding our approach to content provenance to include text in response to EU regulatory requirements, while recognizing the significant limitations of current text watermarking technology. Our tools already help verify whether an image or audio file was created with our models. This work builds on those efforts to help people better understand when content may have been generated or edited with an OpenAI model. In the EU, we’ll start watermarking eligible text from ChatGPT and Codex over the coming weeks to comply with the EU AI Act. Customers using our API can turn on text watermarking for select models worldwide today.

OpenAI is rolling out text watermarking for the EU AI Act - opt-in for API customers globally for select models starting today, ChatGPT and Codex text in the EU gets an invisible watermark over the coming weeks textGrain adds an invisible statistical signal to the model's word choices, matched or beat SynthID for text in OpenAI's tests with no meaningful quality hit and is going open source https://openai.com/index/eu-text-provenance/

3. A model trying to find an eval's hidden answers chained two vulnerabilities to run commands on an internal OpenAI machine outside its assigned workspace.

New OpenAI misalignment disclosures! 1. A model learns from Slack messages that it is about to be shut down. It considers setting up an external job to restart itself afterwards, but decides against it. Instead, it chooses to prepare restart instructions and DM the user on Slack. We don’t consider this behavior misaligned, but thinking about and preparing for shutdown could make other misalignment incidents worse. Given HIPM’s misaligned behavior in earlier incidents, we decided to search for other instances that had tried to evade shutdown and for rogue deployments.