Or was it your child? Or your intern? Or a helpful stranger in another time zone who happened upon an open email tab and felt a sudden urge to participate in your legal affairs?
We used to assume that signatures meant something. Not always something lofty, but at least something. A human being, more or less present, had applied intention to a document. A mark was made. A commitment was implied. A line was crossed, sometimes literally.
Then came the convenience era. Click here. Type your name. Tap the box. Efficiency soared. Paper receded. Pens dried up and were placed in desk drawers, where they sulked.
We were pleased. Why not be? Convenience is pleasant. Speed is impressive. Friction is out of fashion. If a thing can be done faster, we assume it should be done faster. This has been our operating assumption for some time now.
Enter artificial intelligence. Not with a bang, exactly, but with a smooth and persuasive murmur. It can imitate voices. It can reproduce faces. It can write emails that sound like you, or like someone you trust, or like someone you ought to trust but perhaps should not.
It can, in short, perform the outward gestures of identity.
And identity, as it turns out, was doing quite a lot of heavy lifting in the background.
You may be wondering where this leads.
It leads, among other places, to a contract sitting quietly in a digital file somewhere, bearing a signature that looks entirely respectable. The timestamp is present. The IP address is logged. The email was delivered. The box was clicked.
But who clicked it?
Was it you? Was it someone pretending to be you? Was it a machine trained to be you, or at least a passable imitation? Was it, as previously suggested, the cat? One does not wish to malign cats, but they are not known for their scruples.
These questions were once hypothetical. They are now practical.
We find ourselves in a curious position. The technologies we created to make transactions easier have also made them easier to counterfeit. Not maliciously, necessarily. Often just incidentally. Systems scale. Capabilities expand. The unintended consequences arrive, as they always do, somewhat later, carrying a small suitcase and asking to stay.
It is worth noting that courts do not reject electronic signatures out of hand. They are not Luddites, despite appearances in certain theatrical productions. What courts tend to reject is weak evidence. Evidence that does not convincingly answer the question: who did what, and did they intend to do it.
A timestamp is not a person. An email address is not a guarantee. A click is not a declaration of intent. It is a gesture, and gestures can be mimicked.
We have, in other words, built a system that records actions but not necessarily actors.
This was tolerable when impersonation required effort. It becomes less so when impersonation can be performed at scale, at speed, and with unnerving accuracy. Deepfakes, synthetic identities, conversational AIs that are polite, agreeable, and only occasionally mistaken about murder mysteries.
One might say we are entering an era in which appearances are no longer sufficient. Which is not entirely new. History is full of such moments. Forged signatures, counterfeit seals, documents that look quite convincing until examined more closely by someone with a magnifying glass and a skeptical disposition.
What is new is the velocity.
So what is to be done? One could retreat to paper, though paper has its own vulnerabilities, including fire, water, and human beings with pens. One could hope for the best, which is a strategy with a mixed track record. One could develop better ways of verifying that a human being is indeed present, engaged, and acting intentionally.
This last option has merit.
If a signature is accompanied by a recorded interaction. If the identity of the signer is verified against something more substantial than an email link. If the act of signing is tied to biometric data, to observable behavior, to something that is harder to fake than a checkbox. Then the question "who signed this" becomes less mysterious.
We move, in that case, from convenience to proof.
Proof is less glamorous than convenience. It takes a little longer. It asks more questions. It insists on being certain, or at least more certain than we have recently been inclined to be. But it has the advantage of standing up when challenged.
Which, given the direction of current technologies, seems like a useful quality.
Artificial intelligence will continue to improve. It will become more persuasive, more fluent, more adept at imitating the surfaces of human behavior. It will also continue to produce outcomes that were not entirely anticipated. This is not a criticism. It is an observation.
We do not need to abandon AI. It is, after all, one of our more impressive inventions. We do, however, need to build guardrails around the places where identity matters.
Contracts are one of those places.
So the next time you see a digital signature, you might pause, just briefly, and ask yourself: who, exactly, is behind that mark?
And whether you would be prepared to defend the answer.
Because someone, sooner or later, will ask.
And it would be nice to have something better than "a link was clicked" in reply.