Redact.dev discovered roughly 80 WhatsApp help pages that contained hidden code blocks meant for Meta’s own chatbot. The blocks, labeled “HelpContentGalaxyLLMInclude,” gave the LLM rules such as not giving legal advice or mentioning upcoming features. After Wired reported the find, Meta said a bug caused the instructions to appear publicly and has now removed them.
Why it matters
If such hidden prompts leak, users may receive incomplete or restricted AI support, so fixing the bug helps keep chatbot answers reliable.
