OpenAI commits to zero data retention for enterprise ChatGPT customers
01
Top story
Any firm using ChatGPT for project work has had a reasonable question sitting in the background: where does that text go? OpenAI answered it more formally this week. Enterprise customers can now opt into Zero Data Retention, meaning prompts and responses are deleted after processing and never used to train models. Alongside it, OpenAI previewed something called Private Safety Processing, which lets automated systems scan for misuse patterns across multiple sessions without any OpenAI staff reading the underlying content. The technical detail and full rollout are expected in September. For firms already on enterprise plans, the practical change is that you now have a written commitment to point to when clients or insurers ask how their information is handled inside your AI tools.
02
Quick hits
-
Practice
OpenAI and Anthropic are competing directly on enterprise data privacy
The OpenAI announcement came with a competitive backdrop. TechCrunch reports that Anthropic has been pushing its own enterprise privacy commitments, and OpenAI is responding in kind. Firms choosing between ChatGPT and Claude for sensitive project work now have formal privacy commitments from both providers to compare.
03
New tools & releases
-
ToolingMeta AI now has a standalone Mac app with screen sharing
Meta's AI chatbot is available as a dedicated Mac application that can see whatever window you share with it and answer questions about what's on screen. Someone reviewing a spec, a PDF, or a drawing set could get responses without switching to a browser. It also supports dictation across other Mac apps.
Only three candidates qualified today, and two covered the same OpenAI story. The brief runs short as a result.