0
Applied AI·October 6, 2026·1 min read

OpenAI is adding an invisible watermark to AI-generated text

Share

Text watermarks at the model level are a step toward provenance infrastructure that regulators and enterprises have been asking for. If your workflows rely on distinguishing human from AI text — education, legal, compliance, marketplaces — assume provenance signals will become available and start designing around how you’ll store, verify, and act on them.

Applied AI

Analysis: insurers brace for multimillion-dollar claims caused by rogue AI agents, amid concern that execs like Sam Altman and Dario Amodei could be held liable

Once insurers start pricing the risk of ‘rogue’ AI agents and personal liability for executives, AI governance moves from policy deck to balance sheet. If you're deploying agentic systems, review D&O coverage, incident response, and human override mechanisms now — underwriters will soon be asking detailed questions.