Apple capped engineers at a $200 daily AI budget

Apple reportedly cut some engineers’ daily AI spending limit from $300 to $200, a move presented as evidence that generative AI costs are becoming a serious constraint even for wealthy companies. The central argument is that unlimited model use can undermine the economics of products, making cost controls a competitive advantage: companies should cache results, use smaller models where possible, and optimize prompts rather than rely on brute-force inference. Research from the University of Michigan adds an equity and infrastructure concern: identical prompts in non-English languages can use up to 179 times more energy than English prompts and may produce less accurate answers, complicating the cost and quality of global AI services.

Other developments highlight the risks and governance questions accompanying AI adoption. Palo Alto Networks is reportedly negotiating a roughly $500 million purchase of Clutch, whose tools control AI agents’ access to company systems. OpenAI introduced GPT-6 Cyber, a vulnerability-finding and exploit-building model limited to approved security testers. Proton scrapped plans to re-interview staff about their AI use after employee resistance, illustrating the limits of policing adoption through scrutiny. Google’s new school controls have raised questions about whether institutions can retain conventional Search while disabling AI features. The NFL and Carnegie Mellon are also collaborating to analyze football strategy and player performance using game footage and tracking data, underscoring how exclusive datasets can underpin specialized AI tools.