OpenAI pushed an update to its ChatGPT mobile app this week, and buried beneath the usual chatter about faster responses is a feature that should make every American pause: the assistant can now see your screen.
A lot of us are starting to wonder if convenience is just the word we use when we've stopped asking questions.
You grant permission, point your camera or share your screen, and ChatGPT reads what's there — your texts, your banking app, that half-finished email to your boss.
Helpful if you're lost in a settings menu.
Also helpful if you're building a quiet little dossier on yourself without noticing.
We just spent a decade teaching people not to click strange links, and now the most trusted app on the phone is asking for the keys to everything on it.
The update rolled out quietly, the way most of these things do — no alarm, no headline, just a new button and a permission screen most of us will tap through on autopilot.
This is the part where someone says I'm overreacting.
But consider what we've already normalized.
Stack them together and the average American household now leaks more personal data before breakfast than a private investigator could have gathered in a month in 1995.
That's the part nobody in Silicon Valley wants to sit with.
Kids are the fastest adopters of these apps and the least equipped to think about what a screen-reading assistant means for their privacy.
A teenager asking ChatGPT to help with homework is now one permission toggle away from handing over their entire camera roll.
None of this is a reason to smash your phone.
The tool is genuinely useful, and refusing to understand it isn't a moral stance, it's just noise.
But there's a difference between using a tool and being used by one, and that line gets blurrier every time an update ships with a friendly name and a long list of new eyes.
The uncomfortable truth is that we're not really debating features anymore.
We're debating what we're willing to trade for a little friction removed from our day.
And right now, most of us are trading away a lot more than we'd admit if someone asked us directly.
My take: read the permission screen this time.
The update isn't evil, and the people building it probably mean well.
Final Thoughts
But a society that stops asking what a tool can see has already answered the only question that mattered.