
What happened
JPMorgan launched $2,000 monthly spending limits for some engineers using Claude Code, with internal error codes showing "ExceededBudget" and "Budget=2000.0." Users can request increases.
Why it matters
The cap marks the bank's first known AI spending or token limit, suggesting JPMorgan is tying ballooning AI costs to measurable value as it rolls out a more restricted coding architecture.
What to watch
The limit applies only to certain engineers; the Devspace rollout is new and it is unclear when or if all Claude users will move to it, with the Teams group at about 1,900 members.
WHO IT HITSJPMorgan software engineers using Claude Code now face monthly budget caps and a controlled sandbox instead of desktop access. Other banks weighing AI coding costs and security may look to this as a template.
Summaries like this, in your inbox every morning.
JPMorgan's move comes after months of carefully tracking engineers' AI usage. In June, Zachery Anderson, chief data and analytics officer in payments, said at a tech event that some employees were "spending more on tokens than their salary," though there was no firmwide effort to scale back usage. On the bank's second-quarter earnings call in July, CFO Jeremy Barnum said token expenses were at that point "trivial" but that the bank was forecasting meaningful acceleration in the second half of the year.
The $2,000 limit arrives alongside Devspace, a new coding architecture that moves some developers from running Claude directly on their desktops to a contained server environment. In April, Pat Opet, JPMorgan's global chief information security officer, said it is risky to put powerful AI tools on the same desktops that hold emails and sensitive files, and described a goal of agents having "no entitlements." JPMorgan declined to comment on the specifics of token limits or its cost strategy, but a spokesperson said the bank approaches AI with discipline and ties costs to measurable business value.
The test for JPMorgan is whether this controlled architecture can scale across its engineering ranks without slowing the work that has made it a first mover in AI on Wall Street. Much remains undecided, including when or if all Claude-using engineers will move fully into Devspace.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
OpenAI launched Astra for Law, wrapping GPT-6 Astra in a legal search index covering US case law, statutes, re…
Google Labs opened its experimental AI agent CC to households of up to six people
Anthropic detailed three metrics — AI-led R&D, oversight of autonomous AI agents, and compute allocation — dis…
On September 16, NVIDIA, Google and Emerald AI launched the AI Energy Management Alliance (AEMA), joined by 18…

Shiseido Japan's AI agent for ingredient discovery cut search time by 95% and increased proposed ingredient ca…

Anthropic published a blog post on September 8, 2026 (US time), outlining six common prompting anti-patterns t…
