Article deleted but this covers the same https://arstechnica.com/ai/2026/05/energy-supplier-abandons-lake-tahoe-residents-to-serve-data-centers/

If they expect an AIs response to hold up as evidence in court they going to be disappointed. It could be correct or just bullshit that’s a logical consequence of the prompt given.

The next time their targeting hits a school, they will be able to blame the AI so no one has to take responsibility. Just remove all humans from the loop. Autonomous fighting at AI speed and other lies.

Or more likely a couple of months later they encounter the problem again and ask why it wasn’t fixed last time. Ignoring the information or troubleshooting requested. Which is then resent. Some months later they decide to just live with the problem.