The Morning Download: Anthropic Courts Financial Services

Good morning. The growing capabilities of Anthropic’s Claude Code have transformed the work of software development, where previously unheard-of levels of automation are now the norm. On Tuesday, Anthropic said AI is about to transform financial services in similar fashion.

The company announced 10 new AI agents that help automate what it described as the most common forms of financial work, among them, building pitchbooks, closing the books and drafting credit memos, the WSJ Leadership Institute's Belle Lin reported.

Earlier this week, Anthropic partnered with Fidelity National Information Services to develop AI-driven software that would help banks police accounts for signs of financial crimes. The company also unveiled a $1.5 billion joint venture with Wall Street firms to sell AI tools to companies—including those backed by private equity.

Also on Tuesday, Anthropic made its flagship AI product, Claude, easier to use in Microsoft’s 365 business software—which is commonly used by financial firms—and expanded its technical partnerships with Dun & Bradstreet, Moody’s and other data and financial platforms.

Anthropic’s initial foray into the sector came last summer, with the launch of a dedicated vertical offering called Claude for Financial Services. Among its customers are Goldman Sachs, Citadel, Citi and AIG, the company said Tuesday.

Anthropic’s latest moves are designed to “bridge the gap” between how quickly AI is developing and financial firms’ ability to actually use the technology, said Jonathan Pelosi, Anthropic’s head of financial services. “This is the difference between, ‘We’re using AI to help write better emails or do some basic research,’ to an investment bank pitch-building.”

At a New York event for financial services Tuesday, Anthropic CEO Dario Amodei said the company’s ability to commercialize technology and gain revenue is held back not by the power of the models themselves, but by their diffusion throughout the world.

“The technology itself, as I said, is on this exponential, and kind of moves at lightning pace. But large enterprises aren’t like that,” Amodei said.

Accelerating the adoption of new technology inside of a large corporation is easier said than done, however. Companies are bound by layers of governance, as well as bureaucracy, older generations of technology, and people who may feel unnerved or threatened by innovations.

AIG CEO Peter Zaffino, speaking on a panel at Tuesday’s Anthropic event, shared his approaches for breaking through those barriers to AI adoption. It can mean moving beyond the traditional proof-of-concept process to help the company better keep up with the pace of technological change.

The PoC has long been the initial stage of technological exploration inside of large organizations. It’s a small-scale, low-risk trial run in which the point is simply to understand what’s feasible with a new technology and what it might take to scale it across an entire organization. The problem is that the lessons gleaned from that intentional process are often out of date by the time it’s over.

“As Marco said, 10 years is compressed into one, and it's unbelievable how fast things are evolving,” Zaffino said, referring to an earlier remark on the panel by Goldman Sachs Chief Information Officer Marco Argenti.

Working closely with Anthropic, AIG quickly applied Claude to the core of the business, underwriting.

‘We went to the heart of the business with underwriting. Working closely with you (Anthropic) in the …  development of Claude’s capabilities, and our ability to harness that change,” Zaffino said.

There are nonetheless three elements that go into adopting new technology, he said.

Companies need to understand those technologies, no matter where they are looking, and develop good pattern recognition in applying new technologies. “It’s not a tool off the shelf,” Zaffino said. “And the third is the pace of execution. You actually do it,” Zaffino said.

He doesn’t insist on consensus, either.

“I say a lot of times, if you have consensus around the table, it probably means you're not pushing things too hard,” Zaffino said. “I don't vote around my table, but if I did it would be 13 to 1 against me in terms of what we should be doing, but we still did it,” he said.

AI is forcing CEOs to make a stark choice. The WSJ reports how bosses are splitting into two camps over what the technology’s best, immediate benefits are: A tool to cut workers or to make them do more. 

Cut them. In staff memo Tuesday Coinbase CEO Brian Armstrong acknowledged the “down market” as a factor driving cuts, impacting some 700 people at the crypto firm, but noted that restructuring will emphasize AI in operations.“This is a new way of working, and we need to leverage AI across every facet of our jobs,” Armstrong wrote.

Make them do more. Axon Enterprise President Josh Isner recently assured the Taser maker’s more than 5,000 staffers that AI won’t trigger layoffs anytime soon. “I am thinking of AI as the thing that allows our teams to do more, not the thing that replaces our teams,” Isner wrote. 

• See the corporate layoffs so far in 2026: Private-sector job cuts were down 1% in the first quarter, but AI led to 40% more layoffs in tech.

Outside the realm of corporate work, AI is not necessarily welcomed as a panacea.

Freelance copywriters are struggling to stay afloat amid a sea of AI-generated writing. To make matters worse, many are also batting back accusations by arm-chair detectors that AI is penning their copy. “It’s like the new McCarthyism,” one Brooklyn-based copywriter tells the Journal. “It’s just crazy. People are demanding proof of something that can’t be proven.”

Seeking several hundred millions. Since rattling Silicon Valley and Wall Street early last year with a powerful model, DeepSeek has become a key component of China’s plan to have top-class homegrown companies in a range of AI fields. Now the Hangzhou-based company is raising money from government-backed investors, the Journal reports. Some prospective investors have valued DeepSeek at around $50 billion in recent talks, people familiar with the matter said. 

Big Tech agrees to share early models with U.S. Google, Microsoft, and xAI have agreed to share early-stage AI models—with safeguards significantly reduced or removed—with the Commerce Department's Center for AI Standards and Innovation.

The Journal reports that the government will use these versions to evaluate national security capabilities and risks, potentially shaping a new review process for AI tools that pose cybersecurity threats.

Claws for the masses. Google and Meta are preparing to roll out AI agents for consumers inspired by OpenClaw, the open-source project that sparked a wave of Silicon Valley interest earlier this year. Business Insider reports that Google may unveil a personal agent called Remy at its upcoming I/O event. Meta is similarly working on its own OpenClaw agent designed for a range of personal tasks, the Information reports. 

One question worth asking: Does life outside of work really need productivity AI agents?

Chipmakers this week continued to signal accelerating demand from the ongoing AI infrastructure buildout.

The $1 Trillion Club welcomed a new member. On Wednesday, shares jumped 13% in early trading, bringing Samsung’s market capitalization to 1,530 trillion won, equivalent to $1.04 trillion. 

Advanced Micro Devices posted a first-