Claude's refusal is getting stricter and annoying each day
retained text
Is anyone else getting tired of this? My requests worked fine from my Mac. I moved the same thing to a cloud machine and got refused instantly, on the very first request. So I asked the assistant for the obvious workaround: a proxy. It flat out refused. See the screenshot: “I wouldn’t use a proxy, and I won’t build one.” It admits it can’t even confirm there’s a restriction, because the provider doesn’t publish how the limit works. It also admits it hasn’t read the terms. So it’s refusing over a limit it can’t confirm and terms it hasn’t read, and then it warns me I could lose my access. I’m paying for this tool, it’s my account, and I’m the one taking the risk. If I want to make that call, I should be able to. Tell me the risk once and then do the work. Instead I lost time arguing with my own tool and I’m still stuck. The best part is “we go around nothing and find a proper route instead.” Great, except it didn’t give me one. Just a refusal and a nice paragraph about principles. Has anyone found a way to get these assistants to stop moralizing, or do you just do this part by hand?
document fetched 2026-11-03 00:22 · canonical https://reddit.com/r/ClaudeAI/comments/1xlz6m8/claudes_refusal_is_getting_stricter_and_annoying