**[Amazon Insider Reveals AI Cost Overruns, Small Errors Lead to Million-Dollar Bills]** According to monitoring by Dongcha Beating and reports from the *Financial Times* citing informed sources, Amazon has internally disclosed multiple cases of severe budget overruns in AI projects. The most extreme case involved using Claude Sonnet to match author information to product pages. The project ultimately failed but ended up costing $1.8 million—860% over budget—and was only discovered five months later. Another financial auditing tool project accidentally overspent by approximately $541,000. A project aimed at optimizing logistics delivery speed also mistakenly incurred an additional $134,000, which was discovered more than two weeks later. Amazon engineers attributed the issues to deployment errors and a lack of cost controls. Small mistakes that would cost almost nothing in traditional systems can result in "catastrophically expensive" bills in AI projects. The company is currently developing automated measures to prevent similar projects from spiraling out of control. Amazon responded by stating that these cases involve only a small number of teams and do not represent the overall use of AI within the company. Previously, Amazon also shut down its internal AI usage leaderboard because employees, in an effort to climb the rankings, had agents perform meaningless tasks, consuming a significant amount of extra tokens. [Original Article Link]