Who made you
Executive summary
The assistant in question is a Meta AI product developed by Meta (formerly Facebook), created and branded as Meta AI and built on Meta’s Llama family of large language models (LLMs) according to Meta’s public-facing pages [1] [2]. Meta presents Meta AI as both a research division and a consumer-facing virtual assistant that is embedded across its apps and available as a standalone app and developer platform [3] [4].
1. Who built it: Meta’s AI organization and lineage
The technology is the work of Meta’s AI organization, which began as Facebook Artificial Intelligence Research (FAIR) and was later rebranded to Meta AI; Wikipedia describes Meta AI as a research division of Meta that was founded in 2013 as FAIR and later renamed following the corporate rebrand [3]. Meta positions its Meta AI team as the group that develops foundational models and agent technologies intended to power assistants and agents across Meta’s products [5].
2. The technical core: Llama family of models
Meta’s consumer-facing assistant and related experiences are explicitly stated to run on Meta’s Llama large language models; Meta’s product pages and home entry say Meta AI is “built on Meta’s latest Llama large language model,” and Meta’s product blog noted the Meta AI app was launched using Llama 4 in announcements about the standalone app [1] [2] [4]. Those model names are how Meta attributes the underlying generative capabilities of the assistant in its public materials [2] [4].
3. Where the assistant appears: apps and integrations
Meta integrates this assistant across its ecosystem: the company advertises Meta AI inside WhatsApp for chat summaries and image generation with privacy-oriented processing, inside Instagram and Messenger for in-DM assistance, and in a standalone Meta AI app that can interact by voice and text [6] [7] [4]. Meta’s marketing materials describe Meta AI as accessible on WhatsApp, Instagram, Facebook and Messenger and claim the assistant can be used for drafting, image generation and conversational tasks [4] [2] [7].
4. How it can be customized: AI Studio and creator tools
Meta offers a developer/creator-facing platform—AI Studio—that lets creators and developers build custom AI characters or “AI extensions” of themselves and deploy them across Meta’s apps, an offering Meta markets as enabling anyone to create AI personas without coding [8] [9] [10]. Reporting and Meta’s pages describe features for setting personality, name, avatar and interaction rules, and positioning AI Studio as a way for creators to automate DMs and scale engagement [9] [10] [8].
5. Research framing and agent work at Meta
Meta frames its work as advancing foundational capabilities—code world models, latent reasoning, grounded agents and other research domains—to produce agents and tools that act both in digital and physical contexts; Meta highlights research goals and the upstreaming of internal agent needs to their foundational models [5]. Meta also publicly touts productized agent features—like Horizon Worlds’ Creator Assistant for automating development tasks—as examples of applied agent work [11].
6. What can’t be concluded from the provided reporting
The provided sources are Meta’s product pages, a Wikipedia summary and third‑party writeups; they establish that Meta AI and Llama models are Meta’s creations and that Meta distributes assistant experiences across its apps [1] [3] [2]. The sources do not, however, supply independent evaluations of model behavior, training data provenance, or third‑party audits; those topics cannot be answered from the material provided here and should not be asserted without other reporting [5] [4].