Who is supposed to catch it, the lab or you?

Read the line under the chat box.

ChatGPT, Claude and Gemini each print one: it can get things wrong, so check what matters. The labs hand the check to you.

And when a company put a chatbot in front of its customers and then blamed the chatbot, the tribunal put it back on the company (Moffatt v. Air Canada, 2024). Whoever acts on the answer answers for it.

That leaves two things to wait for: the labs grading their own work, or regulators catching up. Neither one is in your chat when the answer arrives.

TrustShell puts the check where the answer lands: on this page, in your chat, in your terminal, in your agent. Two checkers that did not write the answer read it, and you see who they were and what they said.

You still decide. The stamp is a second opinion from two other models, not a ruling from an authority.