Coverage from The Hacker News indicates that WhatsApp recently removed hidden instructions embedded within its help center pages that directed its support chatbot to avoid discussing legal actions or regulatory complaints with users.
The hidden text, discovered by data-deletion software firm Redact, was found in the HTML code of 80 help pages. These instructions specifically directed the artificial intelligence chatbot to refrain from mentioning privacy laws, such as the General Data Protection Regulation, or suggesting that users contact data protection authorities when inquiring about account deletion. This occurred despite WhatsApp’s own privacy policy, which explicitly grants European users the right to lodge complaints with supervisory bodies. Another block of text instructed the chatbot to avoid discussing liability or citing terms of service if a user mentioned lawyers or lawsuits. WhatsApp stated the instructions were intended to prevent the AI from providing legal advice and claimed their presence in the public-facing code was due to a technical bug. The company removed the 102 hidden blocks across its help center by October 18. While it remains unclear if the chatbot followed these directives, the incident highlights concerns regarding transparency in AI-driven customer support systems and the potential for internal instructions to contradict established privacy policies and legal rights.
Source: The Hacker News
