Meta AI is available across various Meta platforms, including Facebook, Instagram, and WhatsApp. According to the company, it’s made some improvements to the chatbot, including its ability to “remember” information from previous conversations. The feature allows the AI to respond in a more customized and personalized way for each user.
For users in the U.S. and Canada, the chatbot will store information about them, such as hobbies or interests. The chatbot will also analyze message content and decide what details will likely be relevant to the user in the future.
“Meta AI will start to give you answers based on what preferences and information you’ve shared,” Meta CEO Mark Zuckerberg wrote on Facebook. “For example, it’s helped me come up with creative bedtime stories for my daughters, so if I ask it for a new one, it remembers they love mermaids.”
Information shared during group chats will be ignored by Meta AI. At any time, users have the option to delete remembered details as well.
No Opt-Out for Meta AI Data Collection
The AI chatbot can access and use information across various Meta platforms. For instance, a music video viewed on Instagram cross-referenced with the user’s home location via a Facebook profile could trigger Meta AI to recommend an upcoming local concert.
Per TechCrunch, there is no opt-out option for Meta AI’s cross-platform data collection. Yet, given how vulnerable the tech can be, many users may be hesitant to trust Meta AI with personal information, and an opt-out choice may eventually appear.
A report from security firm Oligo uncovered a security flaw in Meta’s large language model (LLM) known as Llama. The weakness, discovered in September 2024, left the door open for knowledgeable hackers to wreak havoc on Meta’s systems, such as releasing dangerous malware. However, the problem was resolved before any harm was done.
Introduced in 2023, Meta AI was added to Facebook, Messenger, and Instagram to help with text translation, generate images, and search the internet. The chatbot currently uses Meta’s Llama 3.2, which can process both image and text prompts. Meta’s upgraded model, Llama 4, is due out later this year, and Meta AI will likely utilize the new technology. Hopefully, any possible security flaws will be addressed before the rollout.
