I don’t see how we could possibly prove that the person who clicked “I agree” is still the person using the computer.
Or that any actual human is aware that an agreement was made (since an AI can find a checkbox nowadays or software can be configured to bypass it). One way to add balance could be to require people asking for contracts to actually treat them like real serious legal documents, show up for the signing, and figure out who they are making an agreement with.
That doesn’t matter. If you authorize an agent—human or mechanical—to enter into agreements on your behalf (even by mistake), and the agent presents itself as operating on your behalf, the agent’s decisions will be treated as though they were your own.
Prinicipal-agent law predates computers by a very long time.