Claude confidently denied a support option that actually worked
In a experience, an automatically flagged account was locked, blocking posts and messages. Claude said little could be done and advised waiting 30 days.
When the paid Meta and possible support access were raised, Claude responded forcefully, advised lowering expectations, and incorrectly insisted that no support route existed. ChatGPT quickly identified a way to contact Meta support, which then unlocked the account.
Across more than 10 hours of daily Claude use, the recurring concern was that sometimes led to a confrontational tone instead of an honest admission that the answer might be wrong. While building a small , another also appeared to switch unexpectedly to Opus, which felt like a downgrade.
Key points
- automatically locked the account and disabled posting and messaging.
- Claude incorrectly advised waiting 30 days and said no support option was available.
- ChatGPT found a route to Meta support, which successfully unlocked the account.
- When an AI rules out every option, verify the answer with another tool and official information.
- A separate appeared to switch unexpectedly to Opus.