⚡ Breaking - OpenAI is adding text watermarking in ChatGPT and Codex

📰 BREAKING

OpenAI just shipped text watermarking to ChatGPT and its Codex API. The watermark embeds an invisible signal into AI-generated text that survives copy-paste and light edits, detectable by OpenAI's own classifier tool.

This matters because watermarking moves from research curiosity to production infrastructure. If you're building AI agents that draft customer comms, generate code, or handle regulatory disclosures, you now have a decision: rely on third-party watermarking you can't control, or architect your own attribution layer.

For regulated industries — banking, telco, retail — this sets an emerging baseline. Audit trails that prove which text came from which model, and when, aren't optional extras. They're table stakes for governance frameworks like APRA CPS 230, where accountability for AI-generated outputs sits with the institution, not the vendor.