Anthropic rolls out new Claude financial agents as OpenAI rivalry escalates

San Francisco-based Anthropic, which makes the Claude AI assistant that rivals OpenAI’s widely used ChatGPT, is launching a slate of new AI agent offerings for banks, investment managers, and insurers.

The agents are meant to help those companies’ employees build presentations, synthesize webs of data sets and filings, build financial models, and complete other tasks, Anthropic said on Tuesday.

Anthropic announced the new agents one day after it said it would partner with Blackstone, Goldman Sachs, and private-equity firm Hellman & Friedman to create a new, unnamed entity for expanding Claude across midsize companies. OpenAI is forming a similar venture, Bloomberg News reported on Monday, underscoring stiff competition.

Anthropic, run by the former OpenAI executive Dario Amodei, has muscled further into financial services since it launched a suite of tools catering to banks, financial-technology firms, and others last year.

The so-called agents have pressured software stocks, whose businesses, investors worry, could be threatened by the AI-driven tools. In early February, for instance, shares of companies that sell software to the legal industry fell sharply after Anthropic said a Claude plug-in could tackle tasks such as reviewing legal documents and contracts.

Sam Altman’s OpenAI has its own set of tools for Wall Street, a realm where tech providers are competing against each other for market share. The new agents Anthropic is rolling out Tuesday include a corporate earnings reviewer, a financial statement auditor, and a screener for documents tied to customer due diligence processes.

Claude can now work across Microsoft’s popular workplace applications, like Outlook, and connect with new data providers, Anthropic said.

The capabilities that Anthropic and OpenAI are touting for financial clients come as both start-ups plan initial public offerings as soon as this year. OpenAI closed its latest funding round in March at a valuation of $852 billion, while Anthropic was last valued at $380 billion, according to the companies.

IPOs from either of the start-ups, which have defined the modern AI boom, would be a boon for Wall Street firms supporting them now. Investment banks leading their IPOs would be in line for fees, while funds that backed them could expect their own big paydays.

AI models’ rapid spread has prompted concerns of privacy, accuracy, and security. Anthropic has invested in reinforcement learning specific to finance topics, says Nicholas Lin, head of product for financial services at Anthropic, referring to a way to train agents to optimize their results.

That training doesn’t include any of the firms’ client data, Lin said.

“It’s important for us to build the model capabilities for the next six months, not today,” Lin said in an interview, adding that clients’ input is crucial along the way. “This is how we really shape the exponential, as we call it. Once Claude is able to do something—even if not well—we’re able to get feedback on how good it is at doing that task, and then we bring that back to the model.”

Claude Mythos Preview—a model Anthropic says is so powerful that it is only being released to a select group of companies including Apple and JPMorgan Chase—is “fairly decoupled” from the finance-related offerings that Anthropic announced on Tuesday, Lin said.

Banks and asset managers have launched their own AI agents, some in collaboration with tech providers. Last week, Citigroup worked with Alphabet’s Google to launch an AI avatar that it hopes can improve financial advisors’ productivity and help clients around the clock.

Write to Rebecca Ungarino at rebecca.ungarino@barrons.com