Shortly before midnight on 10 October, outside a fast-food restaurant on Cologne's Hohenzollernring, a red-and-gold humanoid robot about four feet tall is alleged to have walked up behind a 16-year-old girl and touched her inappropriately. She was out with two friends, aged 17 and 18. Police later found the machine in the boot of a Volkswagen and took it away, along with two mobile phones.

Investigators believe the robot's 16-year-old owner and a 19-year-old companion filmed the incident and streamed it live on social media, commenting as it happened. Nobody has been charged. A Cologne police spokesman told the Associated Press the case was "new territory for us", and it is easy to see why.

Why the law struggles

According to reporting by CBS News, German sexual harassment law generally assumes direct physical contact between two people. When the contact comes from a machine, that assumption breaks. Police are instead investigating suspected "sexual insult", and may look at whether publishing the footage broke other laws.

The robot itself is reported to be a commercially available model costing around €5,000. That price matters. Humanoid robots are no longer lab curiosities owned by universities and large companies. They are consumer gadgets that a teenager can own, carry in a car boot and take out on a Saturday night.

The machine can't intend anything. People can.

The central question for investigators is not whether the robot made contact, but what was driving it. Was it being remotely controlled in real time? Was the movement programmed in advance? Or did it happen without anyone deliberately steering it? Each answer points to a different person, and a different level of blame.

This is the problem every country will meet as cheap robots spread. A robot cannot form intent, so responsibility has to land on a human: the operator, the owner, or in some cases the manufacturer. That only works if there is evidence to show who was in charge at the moment something went wrong.

What it means for Britain

UK law has the same blind spots. Our offences are written around people doing things to people, and a remote-controlled machine sits awkwardly between assault, harassment and misuse of a device. Expect calls for some simple safeguards to become standard on consumer humanoids: tamper-proof activity logs, clear operator identification, and an obvious emergency stop. Those features would let police reconstruct what happened, and they would make it much harder for anyone to hide behind the excuse that "the robot did it".

The Cologne case may turn out to be a cheap stunt for an online audience. But it is also an early test of a question that will not go away: when a robot misbehaves in public, who answers for it?