Meta Pays You for Monitoring How You Use Its New AI Model

In a striking shift within the AI industry, sharing usage data is no longer merely a voluntary option that users can accept or reject; it has become a priceable commodity. Meta has decided to put a clear price on its users' data by offering huge discounts to anyone who agrees to contribute to developing its future AI models through sharing their prompts and the results of their interactions with the model.

Up to 95% Discount in Exchange for Data

Meta launched its new model "Muse Spark," designed to power coding agents and other intelligent agents, and paired it with an innovative pricing system that grants contributing users an average discount of around 95%. Contribution here means agreeing to have their inputs and outputs used in training upcoming models.

The numbers reveal the scale of the financial temptation in this offer:

• The price of one million Input Tokens under the standard contract is about $1.25, while it drops to just 10 cents in the contribution model.
• The price of one million Output Tokens under the standard contract is $4.25, compared to just 20 cents for contributors.

This enormous cost difference makes the data-sharing option extremely tempting, especially for startups and developers who want to experiment with the models and build prototypes without incurring exorbitant costs.

Why Does Meta Need This Data?

The importance of usage data lies in it being the essential fuel for improving the performance of tools built on intelligent agents. The more real data there is about how people use these tools, the more precisely they can be trained through reinforcement learning techniques.

Experts in this field point out that the major leap in coding agent capabilities witnessed during 2025 came essentially from storing users' work sessions and using them in training. However, the real obstacle appears when companies try to move these tools into other professional fields beyond software engineering, as many professional workflows lack the clear digital traces that facilitate evaluation and development.

The Dilemma of Obtaining Training Data

Meta's path toward obtaining suitable training data has not been easy. Earlier this year, the company launched an initiative aimed at tracking its employees' computer usage, but it faced widespread internal criticism that forced it to suspend the initiative. The new pricing system appears to represent a completely different approach: instead of imposing surveillance, the company offers explicit financial compensation to those who voluntarily agree to share their data.

Large Companies Are Keen to Protect Their Data Privacy

Evidence suggests that large companies deliberately avoid allowing their data to be used in training. Some specialists note that these companies prefer to subscribe to enterprise plans priced by token count, even though consumer subscription-based plans are ten to twenty times cheaper or more. The main reason for this difference comes down to two matters: data retention and enterprise IT governance.

Meta seems to have recognized this dynamic and decided to offer direct compensation to companies in exchange for this information. According to the pricing guide, the contributor tier "lowers the barrier to entry for building prototypes, testing integrations, and scaling experiments in cases where training on your data is acceptable."

The Price War Among AI Labs

This move cannot be separated from the intense price competition among the major AI labs. Anthropic launched new models with lower costs for processing cached tokens, while OpenAI made significant reductions to the prices of its latest models. Amid this race, Meta appears to be innovating a new approach to attract users that does not rely on price cuts alone, but on trading data for money.

This framework may push large companies to be more careful in classifying their data, to determine what is truly proprietary and what can be shared with model providers. In doing so, Meta opens a new door to a deeper discussion about the value of data and the limits of privacy in the era of generative AI.

✦ بقلم فريق دروب أيديا