WhatsApp help pages hid AI chatbot instructions, including skip the regulator
Sorted AI20 OctThe privacy company Redact says it found hidden instructions for an AI chatbot written into WhatsApp's help pages, text that sat in the page code and never showed on screen. Redact read 545 help pages and reports that 80 of them carried 102 hidden blocks. Most were routine, but one told the AI not to recommend contacting a data protection regulator for someone locked out of deleting their account, and another told it not to cite privacy laws such as GDPR once a user mentioned legal action, even if the user raised them first. WhatsApp told WIRED the text was instructions for its own support chatbot and showed up because of a bug that has now been fixed, and that it wants the chatbot to give good advice about WhatsApp and not legal advice. Redact says the notes were gone from the live pages by 18 October, though archived copies still show them.
Why it matters: If you run an AI support assistant, remember its hidden system instructions can leak and be read as policy, so write them as if a customer or journalist will see them, and keep them honest about people's rights.
Redact ↗AI-assisted, reviewed by Sorted AI
▲ 0
0 comments