JPMorganChase expands Claude beyond engineering under strict controls

By Olivia Hart

Anthropic announced on November 9, 2026 that JPMorganChase is expanding Claude from engineering into product development, infrastructure, risk assessments, and asset and wealth management.

The rollout changes which employees can use generative AI, which systems agents can reach, and who approves new models and production changes. At JPMorganChase, that means staged access, thousands of automated tests, hard pass-or-halt pipeline gates, and continuing human accountability; it does not give autonomous agents authority to approve their own work.

How does Claude Code fit into the expansion?

Claude Code, Anthropic’s coding agent, plans work, writes code, runs tests, and opens pull requests. In the Corporate and Investment Bank, Anthropic says work is moving into production faster, with the resulting capacity redeployed against business priorities. Engineers also are modernizing additional legacy systems and sharing agents, including tools for pull-request review and self-healing.

That extension puts AI-assisted work into several operating cycles. Infrastructure and operations teams use Claude Code for monitoring, alerting, and root-cause analysis. Product teams use Claude to turn requirements directly into engineering backlogs, while other teams apply Claude Code to risk-and-controls assessments and program planning. Research, requirements, design, and planning also fall within the expanded scope.

How are investment teams organized?

In Asset & Wealth Management, JPMorganChase is placing employees with business insight alongside technology and product specialists in small, empowered teams. Claude Code has become part of their engineering workflow. Anthropic says the structure is helping deliver customized client investment solutions at scale and move faster, while every change receives more rigorous testing before clients receive it.

Separately, J.P. Morgan Asset Management announced in October 2026 a connector that brings 55ip’s ActiveTax Technology into Claude for Financial Advisors. The tool lets advisors evaluate the tax impact of moving a client’s holdings into a new portfolio. It combines connectors with workflow skills for research, preparation, and documentation.

Human review remains explicit. Investment recommendations, client communications, compliance determinations, and other regulated activities remain subject to human review and approval. The tax tool therefore supports advisor analysis and documentation, but it does not replace the people responsible for a client decision or compliance judgment.

What controls govern production access?

Claude Code runs inside Devspace, JPMorganChase’s containerized development platform, on the bank’s own infrastructure rather than on developer desktops. Agents do not receive direct access to employee credentials or internal systems. The bank expands access in stages, and each new Claude model must pass JPMorganChase’s approval process before employees can use it.

Thousands of automated tests and approval gates check every change. An enforced pipeline gate returns a hard pass or halt, preventing Claude from approving its own work. People remain accountable for every production change, keeping model output separate from production approval even as agents take on more assigned work.

How does the rollout address security?

JPMorganChase’s security team uses Anthropic’s frontier models to find and fix vulnerabilities faster, according to Anthropic. The company points to Project Glasswing, announced on April 7, 2026, as the earlier description of that work. The launch program called for partners to use Claude Mythos Preview in defensive security work, including JPMorganChase.

At the Glasswing launch, JPMorganChase Chief Information Security Officer Pat Opet described an early opportunity to assess next-generation AI tools for defensive cybersecurity across critical infrastructure. “We will take a rigorous, independent approach to determining how to proceed and where we can help,” he said.

What should employers test before scaling agents?

Anthropic says JPMorganChase’s requirements are shaping what it builds for other regulated institutions. For employers considering similar deployments, the practical tests are concrete: which employees receive access, which systems agents can reach, how models earn approval, and who can authorize production or client-facing changes.

The announcement provides no employee count, adoption rate, or completion timetable. It does define the operating boundary: broader AI participation is paired with infrastructure isolation, staged permissions, model-specific approval, automated testing, and human sign-off. As Claude reaches more financial workflows, those controls will determine how far agents can progress within production processes.