Well, there's your rationale as to why AI cannot replace you.
When sh!t hits the fan, Anthropic will immediately point to this clause. Who knows, maybe a court would see it as valid.
Meanwhile, your customer (and thus, your management) is looking for someone to blame for excrement making contact with the impellers. And that someone's gonna be you.
But you have limited funds to take in a lawsuit realistically the worst they can do is fire you, it's not like being blameable somehow makes you more valuable.
Employees often make mistakes that cost companies thousands of dollars. And there's no shortage of stories where employees cost companies tens of thousands and millions.
When a construction guy messes up measurements and thousands of dollars of work has the be removed and redone, no one thinks of taking the employee to court. Why would you want to take your Ai to court?
Well, OpenAI doesn't seem to have clauses like this. Europeans are allowed to use it for commercial purposes under the ToS. (But check it yourself, I'm not a lawyer).
I reimplemented my startup idea from scratch with Codex a few months ago, just for peace of mind.