Meta adjusts AI evaluation practices amid employee concerns

Meta has revised its internal performance evaluation criteria, removing direct ties to AI tool usage, according to three employees who received the update. The shift comes as the company tests its new agentic AI tool, Hatch, which allows autonomous computer actions. Employees report increased token consumption despite the policy change. Previously, AI usage was linked to labels like 'AI Native,' which some employees argued led to unfair treatment, particularly for those on leave. Meta’s spokesperson, Tracy Clayton, emphasized that evaluations focus on contributions, not AI adoption metrics. Employees say they previously competed to maximize token usage, even creating internal leaderboards. Hatch, similar to OpenClaw, can browse the web and manage applications, but privacy and reliability concerns persist. Some employees now use Hatch for personal tasks, though environmental and economic impacts remain a concern. Meta has paused a previous project that tracked employee keystrokes, which eroded trust in AI tools. While token usage remains high, management still seeks proof of effective AI use. Meta has delayed further layoffs due to AI-related software issues, with CEO Mark Zuckerberg stating no major cuts are expected this year.