Research reveals that around 80 WhatsApp help and FAQ pages inadvertently exposed internal system instructions intended for a Meta chatbot within their source code.
The hidden text included behavioral guidelines and topic restrictions.
According to research by data protection firm Redact, about 80 FAQ pages had hidden instructions for Meta's artificial intelligence system in their code. These texts were not visible to visitors, but contained instructions on how the system should answer and restrictions on topics it should avoid.
What did the secret instructions say?
Among other things, the guidelines explicitly stated that the system should not suggest that users contact data protection authorities or regulators. On pages related to deleting an account, reporting problems, or blocking users, there were instructions not to provide legal advice and not to mention the terms of use or privacy policy, even if the user mentioned possible legal action.
Additionally, the hidden instructions prohibited the system from referring to data protection laws, giving instructions to users who stated they were deaf or hard of hearing without specific precautions, or mentioning features that had not yet been released to the public.
READ ALSO
Nvidia unveils new open AI models for business and computing
Meta's reaction
After WIRED contacted Meta, the texts were removed from all pages.
A WhatsApp spokesperson said the texts were appearing due to a bug that has now been fixed. He explained that the instructions were intended solely for support systems to ensure accurate responses.
G
Hacks.gr on Google
Set it as a preferred source for cybersecurity updates.
Preferred source
Top comments (0)