Meta's AI chat feature on Instagram raises privacy concerns
Meta's new AI personas are a group of characters played by celebrities and creators. Individuals can use it for chatting on platforms like Instagram, Messenger, and WhatsApp. The feature is currently in beta, and ahead of stable release, it has sparked privacy concerns. It seems that conversations with the AI characters on Instagram aren't end-to-end encrypted. When users enable end-to-end encryption, the option to chat with an AI persona vanishes, hinting that these messages might be vulnerable to third-party access.
Data collection and Meta's past reputation
Generative AI's growing popularity has led to more data collection for training models and giving companies in-depth user insights. Meta has faced backlash for its handling of personal data, including the Cambridge Analytica scandal and instances of sharing private conversations with law enforcement. These events raise doubts about whether Meta can be trusted with users' data in AI messaging.
Uncertainty revolves around WhatsApp too
When using the AI messaging feature on WhatsApp, a pop-up disclaimer mentions that Meta may use AI messages to enhance AI quality, but personal messages will stay end-to-end encrypted, per Mashable. However, this statement remains unverified due to Meta's unclear generative AI privacy policy. The policy says that messages sent to AI may be used for training the model, but it doesn't explicitly confirm end-to-end encryption for these chats.
Generative AI services offering opt-out and delete options
OpenAI recently introduced an opt-out feature for ChatGPT, allowing users to prevent their information from being used in training the model. While other chatbots like Google Bard and Microsoft Bing lack opt-out features, they do let users delete their activity. Moreover, Meta's generative AI privacy policy also offers options for users to erase their data by using specific commands in the chat interface.
