WhatsApp AI chatbot received hidden guidelines and rules

Recently, it was discovered that on 80 help pages of WhatsApp, risky and hidden guidelines for the customer service AI chatbot were present. These instructions, which were not visible to regular visitors, included, among other things, that the chatbot should never advise contacting a privacy regulator if a user experienced difficulties deleting their account. WhatsApp has since removed these texts and refers to it as a mistake.

The privacy company Redact conducted a thorough analysis of all 545 pages of the WhatsApp help center and found a total of 102 hidden text blocks. This information was intended solely for the internal functioning of the chatbot and was not meant to be visible to users. The most notable guideline related to account deletion advised the chatbot not to mention contacting a privacy regulator. Redact shared its findings on social media and raised questions about the implications of these instructions.

Moreover, there were more instructions that limited the chatbot in providing information about legal actions and privacy laws, such as the GDPR. When users asked questions about blocking or hacking, the chatbot instead guided them to internal processes without terms of deadlines. However, this policy seems partially intended to protect users, such as the advice not to share incorrect information about birth dates or keeping group chats as evidence.

WhatsApp has confirmed that these guidelines are not related to Meta AI, the platform's general assistant, but specifically for their own support chatbot. The rules have now been removed, an action that followed publicity and inquiries from the media. This incident underscores the importance of transparency in AI assistants and serves as an important lesson for consumers to consult independent information sources regarding privacy issues.

Read the full article from AI Insider.