Meta is bringing a general-purpose AI assistant and a cast of character-based chatbots to its messaging apps. The announcement puts AI tools inside WhatsApp, Instagram, and Messenger, where users can ask questions, plan trips, generate images, or chat with personalities represented by celebrity avatars.
A general assistant built into messaging
Meta AI is designed to handle everyday requests, including answering questions and helping users plan trips. It can also create images from text prompts. During demonstrations, users could trigger image generation with an “/imagine” prompt, and Meta said the feature would be free to use.
The assistant will draw on real-time results from Microsoft’s Bing search engine. That connection is intended to help it respond with current information, alongside its broader conversational and image-generation features.
Meta said its assistant uses a custom model partially based on the company’s LLaMA 2 language model, released in July. The company did not provide full details about how the assistant was trained, leaving the precise makeup of the system unclear.
Celebrity characters give the bots distinct roles
Alongside Meta AI, the company is rolling out 28 AI-powered characters across its messaging platforms. Some are played by celebrities, with each character assigned a recognizable conversational role.
- Tom Brady plays “Bru,” a sports fanatic.
- Kendall Jenner plays “Billie,” the “big sis.”
- Snoop Dogg appears as a “dungeon master” offering gaming advice.
The celebrity images can be manipulated generatively to create new animations, according to TechCrunch. The characters reportedly do not speak, however, and are better understood as visual avatars for text-based chatbot personalities than as talking digital versions of the celebrities.
Meta said the celebrity AI characters entered limited beta on Wednesday. The announcement described their intended roles, but reports of how they perform in everyday conversations had not yet emerged.
Safety work and questions about data
Meta said it spent 6,000 hours on red teaming exercises to identify possible problematic uses of its assistants. The work included generating thousands of internal conversations to help refine how the assistant behaves and responds.
For now, Meta’s assistant has not been trained on public user data from Instagram or Facebook. The Verge reported that the company may consider “social integrations” to make the chatbot more useful, citing Ahmad Al-Dahle, Meta’s VP of generative AI. That possibility points to an open question about how the assistant could connect with the social platforms over time.
Distribution is part of the strategy
Meta’s messaging apps give it a large potential audience for its AI products. Company executives see billions of daily users across WhatsApp, Instagram, and Messenger as an advantage in introducing chatbots to consumers.
The reach of those services could help Meta put an AI assistant in front of people who have not used one before. It may also encourage users to spend more time in Meta’s apps. The launch therefore combines new chatbot features with a broader effort to make AI part of the company’s existing social and messaging experience.
How useful that experience proves will depend on the assistant’s responses, the character bots’ appeal, and the ways Meta connects AI with its platforms. The announcement establishes the range of features and the initial safety work, while leaving those practical questions to be answered as the tools reach users.