Goldman Sachs and Citigroup Are Testing Anthropic's Mythos AI Model Built for Finance
Wall Street's biggest banks are quietly testing artificial intelligence models from Anthropic, according to sources familiar with the matter. Goldman Sachs, Citigroup, and JPMorgan Chase are among the financial institutions piloting Anthropic AI systems internally — a trend already reflected in the surging US business adoption of Anthropic tools, which hit 30.6% of American companies in March 2026.
Which Banks Are Piloting Anthropic AI Models?
Goldman Sachs and Citigroup are leading the charge among major US financial institutions, with sources confirming active pilots of Anthropic models. JPMorgan Chase has been identified as the only bank named in Project Glassway, an internal AI initiative that evaluates model security and responsiveness to cyber attacks. The breadth of participation suggests Wall Street views AI as a competitive necessity, not a luxury.
These pilots go beyond basic chatbot deployments. Banks are testing Anthropic's Claude for complex tasks including trading analysis, compliance review, client communication drafts, and internal knowledge retrieval. Several institutions reportedly encouraged employees to use the AI tools to identify security vulnerabilities in critical financial workflows — echoing concerns raised by the US National Cyber Director's recent push to secure critical infrastructure.
Why Banks Are Choosing Anthropic Over OpenAI
Anthropic's reputation for safety-focused AI development has made it particularly attractive to heavily regulated industries like banking. Financial institutions face strict compliance requirements, and Anthropic's Constitutional AI approach aligns well with those constraints.
The timing coincides with Anthropic's aggressive enterprise push following massive funding rounds and the launch of Claude 3.5 and Claude 3.7 models, which have received strong reviews for complex reasoning tasks central to financial analysis. This enterprise momentum is central to Anthropic's growing rivalry with OpenAI for enterprise dominance — a competition that is intensifying as both companies fight for the same Fortune 500 budgets.
What This Means for Finance and the AI Industry
The banking sector's embrace of Anthropic represents a significant vote of confidence in the startup's enterprise viability. While OpenAI has dominated headlines, Anthropic has been steadily building a customer base among institutions that prioritize reliability and safety. The company has also been deepening its financial sector ties through banking partnerships for its own employees via Mythos.
If major banks formally adopt Anthropic models at scale, it could accelerate enterprise AI adoption across other regulated industries including insurance, healthcare, and legal services. Microsoft-backed OpenAI and Google's Gemini are also competing aggressively for enterprise banking contracts, and the next 12 months could see a major realignment of which AI provider dominates financial services.
Frequently Asked Questions
What is Project Glassway?
Project Glassway is an internal AI security initiative at JPMorgan Chase that evaluates AI model safety and responsiveness to cyber attacks. It is one of several bank programs currently testing Anthropic AI models for enterprise use.
Which banks are testing Anthropic AI?
Goldman Sachs, Citigroup, and JPMorgan Chase are among the major US banks currently piloting Anthropic AI models internally, according to sources familiar with the matter.
Why are banks choosing Anthropic over OpenAI?
Banks prefer Anthropic due to its Constitutional AI safety framework, which aligns with strict regulatory compliance requirements in the financial industry. Claude models also score highly on instruction-following and reduced hallucination rates in structured financial tasks.
The Bottom Line
The quiet AI arms race on Wall Street is heating up. With Goldman Sachs, Citigroup, and JPMorgan Chase all piloting Anthropic models, the financial industry is moving past experimentation and toward real deployment. For Anthropic, cracking open the banking sector would be one of the biggest enterprise wins in the AI industry to date.