JPMorgan Chase has introduced a monthly spending limit of $2,000 per engineer for the use of Anthropic’s Claude Code. This restriction is enforced within the bank's internal secure development environment, known as Devspace, which serves as a controlled gateway for monitoring data flow and maintaining security compliance.
The bank mandated the adoption of approved AI coding tools for approximately 65,000 Global Technology employees in March 2026. While the cap applies to select engineers, the potential financial scale is significant; for 10,000 users, the cost could reach $20 million monthly, totaling $240 million annually.
These measures are part of a broader effort to manage the bank's technology budget, which is projected at approximately $20 billion for 2026. Similar to JPMorgan, Goldman Sachs has also implemented cautious access management for AI tools across its engineering teams to control operational costs.
Additionally, JPMorgan has completely restricted access to Anthropic’s Claude models for employees based in Hong Kong as of June 2026. The bank cited licensing and legal requirements within that jurisdiction as the primary reasons for this total block on the AI tools.
