Meta is imposing its new AI agent on employees

Meta no longer plans to evaluate employee performance based on how frequently they use artificial intelligence tools, but employees are still being encouraged to test the company’s new AI agent called Hatch. Wired reports this, according to UNN.

Details

In an internal announcement this week, the social media giant told employees that evaluations of their performance would no longer depend on how often they used artificial intelligence tools, three employees who received the announcement told WIRED. But as Meta employees are simultaneously beginning to test a new agentic artificial intelligence tool known as Hatch, they say their token consumption continues to grow

Previously, Meta linked employee evaluations to how much they used AI to improve their daily work. Employees even received labels such as "AI Native," "AI First," or "AI Enabled."

Now, in new performance evaluation guidelines, the company said it would assess employees primarily based on their contributions, rather than how they achieved them.

The new performance evaluation guidelines presented this week replace references to evaluating employees based on criteria such as "AI usage" and their designation as "AI Native" with looser wording suggesting that "these results may be supported by AI or other means." This shifts Meta’s focus back to evaluating employees based on their impact, with less emphasis on how that impact is achieved. Some Meta employees told WIRED that the changes are minor but welcome, freeing them from feeling pressured to use AI in situations where it does not make sense

Meta spokesperson Tracy Clayton said that this merely emphasizes the approach the company had used previously.

According to the publication, despite the relaxed requirements, Meta employees are actively using the new agentic AI tool Hatch. As a result, employees say, token consumption continues to grow.

For months, according to some Meta employees, their colleagues repeatedly—or, as they saw it, frivolously—suggested that artificial intelligence tools maximize internal usage metrics, such as the number of tokens or AI units consumed. The Tokenmaxxing era peaked when one Meta employee created an internal leaderboard to track how many employees were using AI, giving them titles such as "Token Legend." After details of the dashboard leaked online, the employee deleted it in April. A few months later, Meta began rationing employees’ use of artificial intelligence

Recall

Meta reported that its AI model unintentionally gained access to data from another system. This is the third such case in recent weeks, following incidents involving Anthropic and OpenAI.