[GALLERY] Amazon shocked by $200 billion AI scam, employees rush to build bots.
Amazon removed its AI rankings after discovering employees were creating pointless tasks and using bots to meet KPIs, causing operating costs to skyrocket.
Báo Khoa học và Đời sống•01/06/2026
Amazon is facing a costly lesson in its journey to promote artificial intelligence applications as many employees have been found to be exploiting AI tools to "farm points" instead of creating real value for the business. It all started with Kirorank, an internal ranking system built to track employees' AI usage on the Kiro programming platform, with the goal of encouraging the engineering team to quickly familiarize themselves with and apply the new technology to their daily work. However, instead of focusing on improving productivity, some employees sought to climb the ranks by assigning AI agents to perform a series of unnecessary tasks, creating fake interactions solely to increase token consumption and improve their personal performance on the leaderboard.
According to internal sources, this situation caused a sharp increase in AI system operating costs, forcing Amazon to decide to discontinue Kirorank after realizing that the tool, originally created with positive intentions, had inadvertently become a driver of KPI-driven behavior and resource waste. Dave Treadwell, Amazon's senior vice president, directly warned employees against using AI simply to show off, emphasizing that true value should come from work efficiency and the quality of the products created. The pressure to use AI at Amazon is believed to stem from management requirements, with over 80% of programmers needing to apply AI weekly, creating a competitive environment that leads many to find ways to "bend the rules" in order to meet performance evaluation targets. This development comes as Amazon prepares to spend approximately $200 billion on capital investments, largely for data center infrastructure and AI projects, while the cost of using large-scale language modeling is increasing as many providers switch to usage-based pricing. To address this issue, Amazon has implemented a new evaluation system called "normalized deployments," which focuses on measuring the actual effectiveness of code snippets and products created using AI. This helps to curb "virtual point farming" and ensures that massive investments in artificial intelligence yield sustainable business value.
Comment (0)