Somewhere in the UK right now, a twelve-year-old is uncapping an eyebrow pencil, drawing a little moustache on their face, and successfully aging themselves past an AI-powered identity check. A third of British kids say bypassing these age gates is easy, and some parents are even helping them do it.

We have built multi-million dollar facial recognition systems, and they are being defeated by a bathroom mirror and a Sharpie. The algorithm looks at a child with a drawn-on moustache and says “checks out.” There is something almost poetic about it. The first mass-scale defeat of AI security is not coming from a nation-state hacking group. It’s coming from kids who found a funny workaround.

The vibe check doesn’t work. That is the real headline.

It’s not just facial recognition, either. “Podslop” is the new term for the flood of AI-generated audio content taking over. Over a recent nine-day window, roughly 39% of new podcasts were likely AI-generated. No human ever spoke those words. We are handing automated systems real-world jobs they are not quite ready for, and then acting surprised when the humans show up and break them in ways the models didn’t see coming. Trust the human signal. It is still the hardest thing to fake.


See why analog markers are currently beating multi-million dollar AI security