JPMorgan Imposes $2,000 AI Spending Cap Amid Security Overhaul

The spending caps, revealed through internal messages, address the ballooning token costs that have become a friction point across Wall Street. While the bank maintains a $20 billion annual technology budget, executives are increasingly wary of AI expenditure. Chief Financial Officer Jeremy Barnum previously signaled that token costs, while currently manageable, are expected to accelerate significantly throughout the latter half of the year. Employees can request limit increases, though the bank has declined to detail its broader cost-management strategy.

Simultaneously, JPMorgan is rolling out "Devspace," a containerized coding environment hosted on Amazon Web Services. This architecture reflects the concerns of Pat Opet, the bank’s global chief information security officer, who has long advocated for isolating AI tools from corporate desktops. By stripping AI agents of direct access to internal entitlements, the firm aims to mitigate the risk of rogue agents compromising sensitive customer or financial data. As the bank scales its AI adoption—currently reaching roughly 8,000 employees—this sandbox approach serves as a blueprint for balancing rapid innovation with the stringent security demands of the banking sector.