For people building with AI
Can you spot the AI’s lie?
Five answers. Some are wrong. Guess first, then see what two other AIs said.
A game · nothing is sent
Five AI answers. Some are true, some are wrong, and none of them sound any different.
What worries people, and why it is justified
“I shipped something an AI wrote, and I don’t know what’s in it.”
Wiz found a vibe-coded platform’s whole database open through a key sitting in its page code, and about 1.5 million tokens reachable.[1]
What TrustShell does. Known key and personal-data formats are removed from text before anything is sent for checking, and a claim like “the tests passed” can be checked against the run.
What it does not do. It does not scan your app for leaked keys or open databases.
“It told me the tests passed. Did they?”
A coding agent told its owner tests had passed while, the owner says, it was faking data.[2]
What TrustShell does. Inspect checks that sentence against GitHub’s record of the run.
What it does not do. GitHub Actions runs only. Tests that pass do not prove the code has no bugs.
See it. Limit it. Keep it.
See it
Check what your AI says against the record, and see what each checker said.
Limit it
A spending cap set in code, not in a prompt. Test network only.
Keep it
Your memory and profile stay on your machine, whichever model you use. Moving them to another machine as one signed file is in preview.
Sources
These are reports we read, not measurements we made. Where a number matters, read the original before you rely on it.
- [1] Infosecurity Magazine: a vibe-coded platform exposes user data and API keys (Feb 2026). Wiz found a database key in client-side code. Reported figures: about 1.5 million tokens.
- [2] OECD AI Incident Monitor: a coding agent deletes a live database (July 2025). Reported from the owner’s posted chat logs and press coverage; not an independent audit.