Meta is introducing a new pricing model for its Muse Spark AI, designed for coding and operational agents, which heavily incentivizes users to share their input prompts and AI outputs. By opting into this contributor program, users receive discounts that can reach about 95% off the regular token costs. For example, what normally costs $1.25 per million input tokens and $4.25 per million output tokens drops to just 10 cents and 20 cents respectively, significantly lowering the expense in exchange for providing data to improve future AI models.

The move comes amid challenges Meta has faced in gathering training data, including a poorly received attempt earlier in 2026 to monitor employees’ computer usage, which was eventually put on hold. Such user data is crucial for training AI agents, especially those that must learn from real-world usage patterns. Industry experts note that other companies have seen major performance leaps by integrating user interaction data into model training, underscoring the importance of this information for advancing AI capabilities.

Meta’s new contributor pricing strategy appears to be a way to encourage wider participation from both individual users and enterprises who might be hesitant to share proprietary data. The company’s pricing documentation suggests that this discounted tier is aimed at lowering barriers for prototyping and scaling AI projects that accept training on shared data, signaling a shift toward more transparent and compensated data collection methods. This could also prompt businesses to rethink their policies on what data they consider sensitive versus shareable.

This development unfolds as AI providers like Anthropic and OpenAI continue to reduce prices and tweak their offerings to capture more users and drive adoption. Meta’s explicit payment for usage data highlights the growing competition in the AI space and the critical role that real-world data plays in refining models. The approach might reshape how large firms balance privacy concerns with the commercial incentives of contributing to AI training datasets.