OpenAI Leak Hints at an Always-On AI That Never Stops Working
An AI that keeps working after you close the app — and may read your inbox.
Something called "o" briefly showed up inside ChatGPT. The tagline attached to it: "your always-on assistant." It appeared for a moment in the list of benefits for ChatGPT Pro, the $100-a-month plan, and then vanished. A separate hidden setting included the phrase "email_suffix: -o," which is the strongest hint that email is somehow involved. The timing is suspicious: OpenAI's DevDay event lands September 29 in San Francisco, the kind of stage where new products get announced. So far, OpenAI has said nothing.
Why does email matter so much here? Because that's where most people's actual work lives. If an AI assistant has its own email identity, it could theoretically draft replies, sort messages, chase follow-ups, or summarize a thread before you sit down. But be careful: the leak does not say any of that. It doesn't confirm the assistant can read your inbox, send messages, forward anything, or manage an account. Right now it's a setting name and nothing more.
The idea isn't as far-fetched as it sounds. OpenAI already sells "agents" — AI that can take actions on your behalf — that run on schedules and use connected tools. The company says its own sales team uses one to combine call notes with account research and draft follow-up emails inside a rep's inbox. A separate product, ChatGPT Work, handles multi-step jobs across files and apps, and OpenAI's newer tools are built to keep running for days without losing track of what they're doing. So "always-on" is a direction OpenAI is already walking.
The catch is simple: leaks are not products. Names change, prices change, features get cut. The email hint could mean an assistant inbox, a routing trick, or something duller. And an assistant that lives in your email would need deep access to your most personal messages — a trade-off of convenience for trust that nobody has been asked to make yet.
- OpenAI is quietly testing something called "o," described in leaked code as an "always-on assistant" — a helper that keeps working after you close the app.
- A hidden setting mentioning email suggests it could connect to your inbox, but OpenAI hasn't said whether it can read, send, or manage anything.
- It briefly appeared as a perk of the $100-a-month ChatGPT Pro plan, hinting at who OpenAI expects to pay for it.
Why It Matters
An AI that watches your inbox and works while you're away could save hours — or demand deep trust.