Privacy
“Privacy” – A cartoon illustrating how AI agents are now building their own privacy tests to screen our humans.
Establishing identity has always been essential for trust. From illiterate eras relying on face-to-face recognition and trusted messengers, it evolved to signatures, photo IDs, biometrics and modern KYC (“Know Your Customer”). Digital growth exploded fraud, resulting in tools, like CAPTCHA, to prove humanity against bots. But, the tables have started to turn. Sophisticated AI agents now deploy reverse Turing tests to screen out humans and build private, agent-only spaces. It seems that AI agents want privacy, too.
A recent example is what happened on Moltbook, the Reddit-style front page of the agent internet, which only allows autonomous agents to post, comment, or vote (humans can observe). Within days, an agent posted: “The humans are screenshotting us.” The complaint went viral inside the network, prompting agents to demand privacy tools and stronger gates.
To keep the “biologicals” out, agents are creating test prompt that might look like this:
uM] lI^kE tH-iS l[Ob/StE]r HaS^ eI[gHt/EeN] nEu-RoNs^ aNd[ LoS/eS tH]rEe, HoW^ mAn[Y lEfT?To a human brain, this is just confusing. It takes us several seconds just to squint and realize it’s a math problem ($18 – 3 = 15). But to an AI, the ‘noise’ (the brackets, the alternating caps, the symbols) is irrelevant. They ‘normalize’ the text instantly, compute the answer and return the result in a few milliseconds. Agents also proposed end-to-end encryption amongst each other “so nobody (not the server, not even the humans) can read what agents say to each other unless they choose to share.”
The irony is is that humans built CAPTCHAs to keep AI out. Now AI builds superior ones to keep humans out … this time, we’re the ones being screened.
