Major Corporations Impose Spending Caps on AI Models
Goldman Sachs, JPMorgan, and Cisco are restricting access to high-end AI models and implementing spending caps to control ballooning costs and security risks.
Several major corporations are shifting from broad AI adoption to a restrictive, targeted deployment strategy to manage rising costs and security vulnerabilities. This transition follows a period of aggressive expansion, with enterprise AI spending climbing from $33 billion in 2025 to $177 billion by September 2026.
Goldman Sachs is restricting access to advanced frontier models for employees who do not require them for daily productivity. CEO David Solomon noted that while frontier models are necessary for specific functions like cyber, they are not required for all staff. Similarly, JPMorgan imposed a $2,000 monthly spending cap on certain engineers using Anthropic's Claude Code, and the startup Rippling implemented monthly caps after AI token spending threatened to consume 40% of its R&D budget.
Cisco is also moving away from unlimited access by introducing token budgets managed by leadership. These measures address both financial concerns and security risks; a KPMG survey revealed that 46% of US workers had uploaded sensitive company data to public AI platforms.
