Google closes an AI Studio delete-flaw report in a minute

Google AI Studio's Delete action doesn't delete anything, according to a researcher's evidence posted September 19. Anyone who typed something in and hit delete, assuming it was gone, was wrong if the claim holds up.

Per the writeup on Hacker News, clicking Delete returns a 404 as if the prompt is gone, but the data stays on Google's backend. Google's AI Vulnerability Reward Program reviewed the report and marked it "Intended Behavior" in about a minute.

That matters for anyone who pasted an API key into a test prompt or dropped in client data for a quick summary, then trusted the UI's word that it was erased.

I want the network traces before I believe the strong version of this. One researcher, no confirmation from Google past the VRP's boilerplate close. "Intended Behavior" is a label Google applies to plenty of reports that turn out to be exactly what they sound like: expected, not exploitable. A screenshot of a 404 doesn't prove the backend still has the row.

The weak version needs no confirmation. Google hasn't published what AI Studio does with deleted prompts. Until it does, the safe assumption is that Delete changes what you see, not what Google keeps. Treat it as a UI state change, not a retention guarantee.

Watch for two things next. Whether Google issues an actual retention statement instead of a VRP disposition label, and whether a second researcher replicates the network capture independently. One report plus a fast bounty close is a claim. Two independent captures is a pattern Google would have to answer.