Meta Forms New Partnerships to Gather Data for AI
Meta forms new partnerships to gather data for AI development, aiming to strengthen model training, performance, and future AI products.
Meta is stepping up its data strategy for AI by moving beyond traditional social posts and into more direct, real-world conversational signals. With fewer people sharing original content on Facebook and Instagram, the company is looking for new ways to gather everyday dialogue, questions, and responses that can feed its large language models.
Why Meta Is Chasing New Conversational Data
Platforms like Reddit and X have become especially valuable to AI projects because they’re rich in natural conversations, Q&A exchanges, and real‑time commentary that map closely to how people use chatbots.
Threads gives Meta some of this texture back, but Reels‑driven, lean‑back consumption doesn’t generate the same depth of text‑based interaction.
To stay competitive, Meta needs more ongoing, fine‑grained insight into how people actually talk, what they ask, and which answers they find useful.
Limitless: Wearable Audio As A Training Stream
Earlier this month, Meta acquired Limitless, an AI wearable startup whose flagship product is a pendant that records conversations and generates summaries and searchable notes.
Limitless has confirmed it will stop selling the pendant and other hardware but will be supported for at least another year for existing users, and its team will shift to helping Meta “build AI-enabled wearables.”
The logical next step is to fold Limitless’s audio capture and summarization technology into Meta’s smart glasses roadmap, turning everyday conversations, meetings and personal notes (where users opt in) into high-quality training and fine‑tuning data.
That would reduce Meta’s reliance on public social posts as a primary language source and give it far more direct examples of how people speak, interrupt, clarify, and negotiate meaning in the real world.
ElevenLabs: Voice, Translation And Localized Reels
Meta has also struck a partnership with ElevenLabs, the AI audio company known for voice cloning and translation. ElevenLabs says the initial focus is on “localizing Reels into various languages,” effectively adding AI dubbing so short‑form videos can reach more audiences without relying on subtitles alone.
The partnership also aims to bring advanced translation tools into Horizon, Meta’s social VR environment, to support more natural cross‑language voice interaction.
Functionally, that means Meta will be processing large volumes of speech, translation pairs, and user interactions across languages, creating a rich dataset of how meaning is expressed, interpreted and re‑expressed in different tongues.
That kind of aligned audio–text data is especially valuable for training multilingual and speech‑native models, and it dovetails neatly with Meta’s ambition to offer “personal superintelligence” on top of its social and hardware stack.
Regulatory Tailwinds From Washington
These moves are also happening against a shifting US policy backdrop. President Trump has issued an executive order framework aimed at “Removing Barriers to American Leadership in Artificial Intelligence.”
This is accompanied with a follow‑on directive titled “Eliminating State Law Obstruction of National AI Policy,” which seeks to preempt state AI regulations that conflict with a national, pro‑innovation agenda.
Draft and final texts emphasize a “minimally burdensome, uniform national policy framework for AI” and direct agencies like the FTC and FCC to challenge state rules that significantly constrain data use or AI deployment.
While the orders do not explicitly mention Meta or wearables, they signal that the federal government is inclined to limit state‑level privacy and AI laws that could restrict data collection seen as important for US AI leadership.
That environment may give large platforms more confidence to experiment with advanced conversational capture, so long as they can argue their practices fit within a national AI strategy.
Bottom Line
Limitless offers a path to ambient, real‑world audio as ElevenLabs opens the door to large‑scale, cross‑language voice data from Reels and Horizon. Either way, it’s clear Meta now sees conversational data not just social engagement metrics but as critical fuel for its next generation of AI models.
