Particle.news

Firms Pull Back on AI as Costs, Poor Outputs and a Security Shock Force a Rethink

Rising licence and compute bills plus error‑prone AI that creates extra verification work have driven companies to cap spending, push emergency patches and press for more local cyber capacity.

Overview

  • Several large firms have started to restrict AI use after token and licence bills rose sharply and tools produced superficial or incorrect results that increased staff verification work.
  • Accenture instituted a per‑employee AI budget of $29 per month after executives said costs 'exploded', illustrating a wider corporate move to limit routine model use pending clearer returns.
  • Surveys of German MINT workers show substantial churn, with about 38.4% reporting colleagues left in the past year for AI‑related reasons because jobs changed, career paths faded or workloads increased.
  • Security researchers used OpenAI and Google models to find a critical PHP vulnerability that forced emergency patching, prompted BSI and Cyberagentur warnings about attackers weaponizing flaws in minutes or hours, and led authorities to advise temporary site takedowns.
  • Banks are raising IT spending while many executives expect automation to cut roles by up to roughly 30% by 2029, prompting calls for better data practices, local compute capacity and research to reduce dependence on foreign AI providers.