German firms curb ChatGPT and Claude use as AI costs skyrocket

Large corporations are tightening access to generative‑AI tools such as OpenAI’s ChatGPT and Anthropic’s Claude after facing rapidly rising expenses and poor output quality. Citi temporarily blocked its staff from the newest Anthropic and OpenAI models, while Atlassian forecasts AI‑related spending of about $120 million a year – roughly three times its prior level. Amazon introduced token limits after scrapping an internal AI‑ranking system, and Adobe warned employees that unrestricted AI usage will soon end. An unnamed media company reached its monthly ChatGPT quota, with half of the usage traced to a single employee who generated little tangible value.

Company leaders report that costly, inaccurate AI‑generated content has led to customer complaints and internal friction. Some CEOs have threatened termination for staff who send uncorrected AI‑written emails, and one tech firm imposed a total ban on AI tools. To regain control, firms are deploying internal dashboards that show real‑time AI spend and issuing recommendation lists that match specific models to particular tasks. The shift marks a move from the earlier push for maximal AI adoption to a more measured, cost‑aware approach.