Genuine question, and I do not think there is consensus yet.
A user tells their assistant: find me a tool that does X, make an account, start the trial. The agent does exactly that. It fills in the form with the user's real details, receives the verification email in the user's real mailbox, clicks the link.
Is that a signup you want?
The case for yes: it is a real person who genuinely wants your product, and the only difference from a normal signup is which fingers typed. Refusing it means refusing a customer for the crime of using a tool. That reflex gets more expensive every month, and it is invisible in your metrics — refused signups do not complain.
The case for no: your terms probably say a human must accept them. Nobody read your onboarding. Nobody saw the pricing page. The person may not even know which product got chosen. And whatever the agent agreed to on their behalf, you may struggle to enforce.
My own position is that the signup is fine and the terms acceptance is the actual unsolved bit — but I hold it loosely.
There is also a version that is straightforwardly not fine: the same assistant, told to open fifty trial accounts. It will honestly declare itself user-initiated every single time, because it is telling the truth. Identity does not catch that; only volume does.
What are you doing in your product right now — allowing, blocking, or has it not come up yet?