Understanding the 'Threat'
The arrival of Meta AI inside WhatsApp marks a significant change for the world's most popular messaging app. Users can now tag '@Meta AI' in group chats or message it directly to ask questions, generate images, or get suggestions. However, unlike your
personal messages, which are protected by end-to-end encryption, your interactions with the AI are a different story. When you send a prompt to Meta AI, that data is processed on Meta's servers. According to the company, this information can be used to improve its AI models and personalize your experience, which may include targeted ads in the future. While Meta states that the AI cannot see your other private chats unless it is specifically tagged, the core issue is that any information shared with the AI is not encrypted in the same way. This creates a new channel through which your data—if you're not careful—could be stored, analyzed, and potentially exposed. The 'threat' isn't about the AI secretly reading all your chats; it's about users unknowingly sharing sensitive information with a system that is designed to learn from it.
Rule 1: Treat the AI Like a Public Forum
The first and most crucial verbal rule is a mental one: 'What I tell the AI, I am saying in public.' Before you ask Meta AI a question or share information with it, pause and consider if you would be comfortable posting that same content on a public website. Avoid sharing any personally identifiable information. This includes your bank account details, credit card numbers, tax information, passwords, or specific health concerns. Even seemingly harmless details, when pieced together, can form a detailed profile. The AI is a tool, not a confidant. Sharing workplace drama, confidential business plans, or intimate personal secrets is risky. These conversations can be stored and reviewed, and while Meta has policies to protect sensitive topics, the safest assumption is that anything you type to the AI could one day be seen by others. By treating the AI as a public space rather than a private journal, you naturally create a strong barrier against oversharing.
Rule 2: Ask 'Is This Real?' Out Loud
The second verbal rule is a habit of verification: ask 'Is this real?' when you receive unexpected or unusual information, even if it seems to come from a known contact within a group chat. The rise of AI tools also fuels more sophisticated scams. Criminals can leverage AI to craft highly convincing messages for 'pig butchering' scams, where they build trust before asking for money or cryptocurrency. If a friend in a group chat suddenly messages everyone with an urgent request for money or a link to an unbelievable investment opportunity, it's worth questioning. An AI could be used to mimic their style. The simple act of sending a separate private message or, better yet, making a quick phone call to verbally confirm the request can thwart a potential scam. This rule also applies to information generated by the AI itself. AI models can 'hallucinate' and present inaccurate or outdated information as fact. Before acting on AI-generated advice, especially for important matters, verify it with a trusted source.
Rule 3: State Your Boundaries in Your Settings
The third verbal rule is a command to yourself: 'I will set my boundaries.' While you can't completely remove Meta AI from WhatsApp, you can take steps to control its influence and manage your privacy. Start by exploring WhatsApp's privacy settings. You can find options related to AI and data usage, though their location can change with app updates. For group chats, familiarize yourself with the 'Advanced Chat Privacy' feature. When enabled for a specific chat, it can block certain AI features, like the ability for another member to summarize the chat using the AI. Group admins can also lock these settings to prevent others from changing them. Recently, Meta announced an 'Incognito Chat' mode for Meta AI, designed to offer a temporary, private conversation that isn't saved or used for training. Seeking out and using these more private modes is a clear way to enforce your data boundaries. The key is to actively manage your settings rather than passively accepting the defaults.
