Release notes add generated content watermarking

The developer release notes for the platform have been revised to document the introduction of a watermarking feature for generated outputs. Prior to this update, the changelog contained no references to embedded identification marks in API responses. With this change, the platform officially acknowledges support for identifying machine-generated material, a capability generally used to prove content provenance or support synthetic text detection.

In practice, developers building software with these endpoints will need to evaluate how marked text might interact with downstream systems. Nonetheless, the brief changelog entry leaves several essential operational specifics undisclosed. It does not explain whether the marking mechanism is applied automatically to all accounts or if it requires an explicit parameter in API requests. Additionally, the update provides no clarity on the mathematical method used—such as statistical distribution shifts during token sampling or cryptographic tagging—nor does it clarify whether the process introduces added latency or affects pricing and token limits. It also remains uncertain whether the feature applies universally across all models or is restricted to specific flagship versions.