Francine McKenna is an adjunct professor of accounting at Montclair State University and also teaches in the University of Cambridge Executive Master of Accounting Program. She is the author of the newsletter The Dig.
FT Alphaville has been trying its best to keep up with the nonstop, increasingly self-reinforcing, multiyear spree of AI ecosystem partnerships, alliances and influxes of investment into each other.
But there’s a lingering question we’ve not yet answered. The answer sheds some light on how key players are lining up to take full advantage of the AI boom, even those we hope will warn us if it’s time to bust.
It’s a simple one: which firm is signing the Anthropic audits?
We can now answer that: it’s probably Ernst & Young.
Anthropic and its bitter rival OpenAI are still private. Both reportedly plan to IPO soon, having announced in June — just a week apart — that they had filed confidential registration statements with the SEC.
Both OpenAI and Anthropic are witnessing SpaceX flapping around like one of those inflatable gas station tube men.
SpaceX’s share price — all Elon Musk-related companies including Tesla are audited by PwC — has been volatile since its IPO. It closed above its $135 IPO price for the first time in nearly a month on Monday.
We do not officially know which Big Four firm — and it has to be one of the biggest ones — audits OpenAI. However, based on FTAV’s reporting, some well-placed sources and a process of elimination based on the auditor independence rules, we’ve known since November of last year that it’s Deloitte.
FTAV was aided in its quest to find OpenAI’s auditor by the UK’s Companies House business register, where OpenAI’s UK subsidiary had named Deloitte Ireland as its auditor. So, we checked for similar clues on Anthropic.
There is a Companies House filing for Anthropic Limited, its UK subsidiary showing Grant Thornton UK as auditor for the year ending December 31, 2024, but nothing more recent.
However, Grant Thornton’s receipt of private equity investment on both sides of the pond — in the UK from Cinven and in the US from New Mountain Capital — complicates things for the firm with regard to independence from Anthropic. Cinven utilises Anthropic’s Claude models through a partnership with AI and data consultancy Artefact and at least one New Mountain Capital portfolio company, Bounteous, is listed as a services partner of Anthropic. Based on auditor independence rules for US-listed companies, that would likely, in my opinion, disqualify Grant Thornton from taking on Anthropic as a US-listed audit client.
Size also matters. Having a Big Four audit remains a stamp of quality for major companies — especially ones who will be targeting valuations that may be in the trillions of dollars.
Which is one of several reasons why we bet the global operation is audited by EY.
Auditor independence rules regarding business alliances are not new. They predate the Sarbanes-Oxley Act of 2002, enacted after the failure of Enron and the implosion of Arthur Andersen, because they are fundamental to maintaining the public’s trust in the integrity of public company financial reporting.
Back in April 2004, when the SEC sanctioned EY for auditor independence violations related to its years-long business relationship with an audit client, software firm PeopleSoft, the judge wrote that:
[D]irect and material indirect business relationships, other than as a consumer in the normal course of business, with a client . . . will adversely affect the accountant’s independence with respect to that client. Such a mutuality or identity of interests with the client would cause the accountant to lose the appearance of objectivity and impartiality in the performance of his audit because the advancement of his interest would, to some extent, be dependent upon the client.
Relationships between an auditor and its client that are prohibited include joint business ventures, limited partnership agreements, investments in supplier or customer companies, leasing interests (except for immaterial landlord–tenant relationships), and sales by the accountant of items other than professional services.
To qualify as OpenAI’s auditor, Deloitte had to plan ahead. That meant abstaining from signing any alliance or partnership agreements to develop and market OpenAI products to its clients during the audit years it would eventually expect to include in the IPO S-1. For example, if the expectation was OpenAI, or Anthropic, had a goal to IPO in 2026, the retained audit firm had to be free of conflicts for 2023–2025, to sign those opinions.
As Louis laid it out back in November:
A more problematic relationship would be if OpenAI was being audited by a firm that is also selling OpenAI products to its clients. Here is a clearer clue. EY and KPMG both sell products that employ Microsoft Azure OpenAI tools (for strategic intelligence and assurance respectively), while PwC earlier this year became OpenAI’s “first resale partner and largest enterprise user”, selling ChatGPT as a complement to their audit and tax services. Deloitte, meanwhile, is definitely using OpenAI internally, but does not seem to be selling it. Instead, it has a business partnership with Anthropic.
Deloitte announced its strategic partnership with Anthropic in October 2025, expanding its deployment of Claude to nearly 500,000 employees worldwide. Described as an alliance and a long-term strategic commitment, Deloitte and Anthropic also plan to build compliance-focused AI products. Last month, Deloitte announced the launch of a platform for secure software developed on Anthropic’s AI technology foundation to help its clients tackle tough cybersecurity problems.
So, Deloitte is conflicted with regard to Anthropic. Deloitte decided a while ago to be OpenAI’s auditor and not Anthropic’s, too. It would have been highly unusual for the fierce rivals to allow the same firm to audit both. Although, as I’ve written before, it’s not unprecedented:
But I have always thought that Uber and Lyft are way too cozy and close and will eventually become one company. Why? Because they are both audited by PwC out of its San Francisco office.
PwC is also doing too much client-facing business with Anthropic to be its auditor. On May 14, 2026, Anthropic and PwC jointly announced they were expanding an existing alliance “driving impact across client work and the firm.” Anthropic said the expansion of their strategic alliance would deepen “how PwC uses Claude to build technology, execute deals, and reinvent enterprise functions for clients across every industry it serves.” Therefore, under current auditor independence rules, PwC can’t be Anthropic’s auditor, either.
KPMG, like Deloitte and PwC, is going all in on Anthropic. In May, KPMG and Anthropic announced KPMG Digital Gateway Powered by Claude, “bringing Anthropic’s frontier AI directly into KPMG’s client delivery platform”. KPMG in the US is also embedding Claude into its PE-focused product offerings, and the two companies plan to co-develop new Claude-powered products for PE portfolio companies.
So, KPMG is out as Anthropic’s auditor. But, hey, the firm recently picked up Tether as a consolation prize. That should be fun!
So, EY is the only Big Four firm eligible, based on current auditor independence rules, to be Anthropic’s auditor, because it is the only one that has not signed a partnership or alliance agreement with Anthropic for client-facing work. The firm mapped out a clear path to becoming Anthropic’s choice for auditor by sticking with Microsoft (and, by extension, OpenAI) for its AI strategy via a $1 billion investment to integrate Copilot and Azure tools for use by its employees and clients.
Neither EY nor Anthropic responded to a request for comment confirming an audit relationship.
Why does this matter? To understand, let’s quickly step back. As the Wall Street Journal’s Asa Fitch wrote in June, it’s significant which of Anthropic or OpenAI goes public first. The first to list might suck all the investor air out of the room to the laggard’s detriment.
I’m not so sure. What I do know is that the first to IPO and publish audited financials will expose more disclosures about the terms and selected accounting for partnerships, joint ventures, investments, and loans that, until now, Microsoft and Amazon, for example, have been very stingy about.
It’s been written a million times before that the AI software and data centre ecosystem is a circle-something… a circular motion machine. Yeah, that’s the ticket! Here, via Morgan Stanley by way of FTAV, is how the concentric circles, money flows, feedback loops between the hyperscalers, chipmakers, and OpenAI looked as of March (several more deals have been struck since then):
A diagram centred around Anthropic would probably show a similar level of complexity. Amid this money morass, we should care deeply about who is checking the numbers.
EY is the dominant audit firm for the hyperscalers, the firms building AI data centres: Amazon, Alphabet, Oracle, and Meta. That adds to the firm’s challenge of making independent, objective, and symmetrical decisions about the correct application of key accounting rules such as fair value, equity method accounting, consolidations, and construction in progress and depreciation when your audit clients are on all sides of investments, partnerships, and alliances.
I’d like to still believe one key actor in all this business is, at least on paper, trying to avoid egregious conflicts of interest. Auditor independence rules prohibit public accountants that sign audit opinions from having partnerships, investments, or business alliances with audit clients for anything other than “ordinary course of business” products or services. That means Microsoft’s auditor Deloitte can load its employees’ laptops with Windows software and Dell’s auditor PwC can issue Dell laptops to its staff, but neither is allowed to sell their audit clients’ products to other clients or financially benefit from promoting them to others.
The same week FTAV revealed OpenAI’s auditor was Deloitte, the FT’s Stephen Foley reported that US Securities and Exchange Commission chief accountant Kurt Hohl announced he’s in favour of loosening the auditor independence rules:
Independence rules that prevent accounting firms auditing companies with which they have a business relationship might not be “fit for purpose” because of the increasingly complex interactions between technology providers, Kurt Hohl, who joined the SEC earlier this year to oversee audit regulation, told a conference in New York. Hohl, a former partner at Big Four accounting firm EY, said the emergence of large AI companies had made enforcing independence rules more difficult.
We won’t know for absolute sure which Big Four firm signs for OpenAI or Anthropic until the SEC finishes its review and their IPO prospectuses are made public. Under current SEC rules we’ll see three years of audited financial statements, with an opinion — hopefully unqualified — signed by one of the Big Four audit firms. We’ll also learn if auditors had any additional concerns about weak internal controls.
But the Big Four global audit firms are an integral part of the tangled web of AI-related investment and commercial activity. There’s really big money in the AI ‘boom’ for the global public accounting firms — whose wider operations also exist in one of the areas most at risk from AI disruption. It could be a make-or-break moment.
