Anthropic deploys Claude across Wall Street: JPMorgan, Goldman, Citi, AIG, and Visa now in production
Anthropic used an invite-only financial-services briefing in New York on May 5 to disclose its largest enterprise wedge to date, debut a finance-tuned Claude Opus 4.7, and present internal benchmarks from JPMorgan and AIG.
Anthropic on May 5 used an invite-only briefing in New York to consolidate a roster of Wall Street customers and a new finance-tuned frontier model into a single product story. JPMorgan Chase, Goldman Sachs, Citi, AIG, and Visa were named as Claude users in production, in addition to a suite of pre-built financial-services agents and the public debut of Claude Opus 4.7.
The new model, Anthropic said, currently leads Vals AI’s Finance Agent benchmark with a 64.4% score and tops the GDPval-AA evaluation for economically valuable knowledge work. Anthropic positioned Opus 4.7 as the most capable model the lab has shipped for financial work to date.
Customer disclosures at the briefing were unusually specific. JPMorgan’s chief executive Jamie Dimon, recounting his initial test of the system, said: “I want to know about asset swaps and Treasury bid-ask spreads, and quitting the markets, and investment grade.” Peter Zafino, the chief executive of AIG, described an internal evaluation: “Claude out of the box scored 88% as accurate as a human expert on insurance claims.” Lori Beer, JPMorgan’s chief information officer, summarized the operating posture: “there’s this capability overhang. The technology can do so much.”
The disclosures matter as a data point on how the largest banks are moving from internal experimentation to production. JPMorgan said its LLM Suite is now used by 150,000 employees weekly across more than 450 AI use cases. Goldman Sachs has rolled an AI assistant to all of its more than 46,000 employees. Citi has put more than 70% of its 182,000 employees on firm-approved AI tools.
Anthropic first launched Claude for Financial Services in July 2025. The May 5 briefing is the first time the lab has put the deployment scale into a single, on-the-record account.
Sources
— END —