Meta shuts down internal Claudeonomics AI tool after employees started leaking data to public

Meta has reportedly shut down its internal AI leaderboard, "Claudeonomics," just days after launch after employee usage data began circulating publicly. The system gamified AI usage, pushing employees to use more AI tools as part of a growing "tokenmaxxing" trend.

In Short

• Meta launched an AI leaderboard tracking employee token usage
• The leaderboard named Claudeonomics ranked top 250 employees by tokens used
• But the company has now reportedly closed the leaderboard after internal data was shared in public

Meta recently introduced an internal AI leaderboard with the aim of making AI a central part of every employee's job. The leaderboard offered insights into how much Meta employees were using generative AI tools by calculating their token consumption. However, within days of its rollout, the company reportedly shut it down, citing concerns that internal usage data was being shared publicly.

According to a report by The Information, the new leaderboard system at Meta—known as “Claudeonomics”—tracked and ranked employees based on their AI token usage. Tokens are units that measure the input and output processed by large language models, effectively indicating how much employees rely on AI tools in their work.

The leaderboard reportedly monitored the top 250 employees across Meta, ranking them by total token consumption. It also gamified the process by assigning badges such as “Token Legend,” “Model Connoisseur,” “Cache Wizard,” and “Session Immortal.” The system was named after Anthropic’s Claude model, which has reportedly been widely used within Meta, particularly for coding-related tasks.

Data cited in the report suggests that Meta employees collectively used around 60 trillion tokens in a span of 30 days. The highest-ranking individual on the leaderboard reportedly consumed 281 billion tokens.

Even as Meta pushed employees to use more AI tools, it has now abruptly shut down the leaderboard, saying it was intended as a fun experiment but was taken down after internal data was shared publicly. “It was meant to be a fun way for people to look at tokens, but due to data from the dashboard being shared externally, we’ve made the decision to shutter Claudeonomics for now,” says the message from Meta. (via Gizmodo).

Meta has not disclosed updated figures on token usage following the removal of the system.

Companies are now tracking how much AI employees are using

Notably, Meta is not alone in encouraging employees to use more AI tools. There is a growing shift across the industry where companies are increasingly tracking employee AI usage and, in some cases, promoting it as a productivity metric. This practice is referred to as “tokenmaxxing,” which refers to maximising the use of AI tools by consuming as many tokens as possible.

In fact, many industry leaders are encouraging companies to evaluate employees based on how much they use AI tools. For instance, Nvidia CEO Jensen Huang has previously indicated that he expects engineers to make extensive use of AI resources.

At the GPU Technology Conference (GTC) 2026 and on the All-In Podcast, he proposed a new compensation model where employees, particularly engineers, receive AI tokens as part of their pay packages. Huang suggested that engineers should receive a token budget worth roughly half of their annual base salary.

For a staff engineer earning $500,000, Huang expects them to consume at least $250,000 in AI tokens annually, treating AI compute as a "fourth pillar" of compensation alongside salary, bonuses, and equity. Huang believes that in future "how many tokens come with this role" will become a standard question in Silicon Valley interviews, similar to asking about stock grants.

This idea aligns with OpenAI CEO Sam Altman's concept of "Universal Basic Compute," where he suggests that access to AI resources should become a fundamental unit of value and potentially a future form of income.

- Ends