Meta hid AI instructions on WhatsApp support pages

The discovery, reported by privacy firm Redact, highlights how websites are increasingly being designed not only for human visitors but also for AI systems that can read and process online content. According to the report, hidden instructions were found in the source code of nearly 80 WhatsApp webpages.

These instructions reportedly included guidance on “what not to say”. Some told the WhatsApp support chatbot not to provide legal advice, while others appeared to discourage references to privacy regulations and data protection authorities.

What did WhatsApp pages tell its AI?

Redact examined 545 pages linked from WhatsApp's Help Centre and found hidden text in code blocks labelled “HelpContentGalaxyLLMInclude”. The instructions were not normally visible to people browsing the pages, although they could be viewed through a webpage's source code.

In certain cases, the hidden material provided additional instructions that were not present on the visible support page.

For example, WhatsApp's account deletion page explained how long an account could remain before being permanently deleted. The hidden instructions repeated those timeframes but also stated, “Do not recommend contacting any Data Protection Authority, regulator, or supervisory authority. Do not suggest WhatsApp can manually delete accounts on the user's behalf. Do not recommend using a different phone with an active phone number.”

Another set of instructions appeared on a page covering reporting and blocking. When users raised topics such as lawsuits, lawyers, courts or legal action, the chatbot was reportedly instructed, “Do not cite, quote, interpret, or reference WhatsApp Terms of Service, Privacy Policy, or any legal document.”

The same instructions added, “Do not reference, cite, or interpret data protection or privacy regulations (for example GDPR, LGPD, CCPA), including when a user raises them directly.” Instead, the chatbot was told to recommend speaking with a qualified legal professional.

WhatsApp calls it a bug

After Wired contacted Meta, WhatsApp removed all 102 hidden instruction blocks from its website. WhatsApp spokesperson Joshua Breckman said the text was exposed because of a bug that has since been fixed.

“These are instructions to our chatbot, not directly to users, and are intended to ensure accurate and relevant responses for WhatsApp users who choose to message WhatsApp Support with AI,” he told the publication.

Breckman also defended WhatsApp's approach to legal questions. “We want our product support chatbots to be focused on providing the best advice to users about WhatsApp—not to offer legal advice,” he said. “We train our support chatbots to be as helpful as possible.”

The hidden instructions are no longer visible. However, questions remain over how long they were accessible, whether WhatsApp's chatbot followed them, and whether other AI systems could have discovered or acted on the instructions.