Technology
Meta introduces steep pricing discounts for AI users who share training data
Users of Meta's new Muse Spark model can receive an average 95% price cut in exchange for allowing their prompts and outputs to train future models.
The short version
- Meta is rolling out a contributor pricing tier for its Muse Spark AI model that slashes rates by roughly 95% for users who consent to sharing usage data.
- The reduced pricing aims to help Meta secure real-world workflow data necessary to improve future agentic AI systems.
- It remains uncertain whether significant numbers of enterprise clients will accept data sharing given ongoing corporate privacy and governance restrictions.
Key facts
- Meta's contributor tier prices input tokens at 10 cents per million compared to $1.25 under standard terms, while output tokens drop from $4.25 to 20 cents per million.[TechCrunch]
- The discounted rates apply to users of the Muse Spark model who agree to share prompts and outputs to assist in future model development.[TechCrunch]
- Meta previously paused an internal employee computer usage tracking initiative in June following workplace criticism regarding training data collection.[TechCrunch]
What remains uncertain
- The extent to which corporate and enterprise clients will participate in the discounted data-sharing tier despite traditional enterprise governance policies remains unknown.[TechCrunch]