AI
WhatsApp help pages hid AI instructions that could suppress privacy guidance
An investigation found 80 hidden AI-directed text blocks across WhatsApp’s help centre, including instructions that could discourage its support chatbot from recommending regulators or citing privacy law after users mentioned legal action. The instructions conflicted with rights described in WhatsApp’s published European privacy policy, although the investigation did not establish that the chatbot followed them. WhatsApp said the notes were intended for its support chatbot. The material was later removed, leaving questions about duration, language coverage, and actual user impact.
Editorial summary · Publisher article
What matters
- Researchers reviewed all 545 pages linked from WhatsApp’s help centre and found 102 hidden blocks.
- Two instructions concerned account deletion and withholding legal or regulatory information.
- The hidden text was transmitted to search-engine crawlers, although the researchers did not test whether external AI systems acted on it.
- The investigation found no comparable notes on 300 reviewed Facebook and Instagram help pages.
What remains unproven
The investigation examined the instructions rather than chatbot answers, so it cannot show whether WhatsApp’s chatbot followed them or whether outside AI assistants obeyed them. It also could not establish exactly how long the notes existed or whether other languages and unlinked pages contained similar material.
Context & caveats
- The findings show hidden instructions and conflicts with published rights, not confirmed instances of users being denied those rights by the chatbot.
Why this made the edition
The investigation documents hidden AI-directed instructions across 545 WhatsApp help pages, compares them with published privacy rights, explains WhatsApp’s response and the limits of what was verified, and identifies important unresolved questions.
Source material