The leaders of the major artificial intelligence companies are due at the White House today to meet US President Donald Trump and House Speaker Mike Johnson, with the oversight of AI expected to top the agenda, an issue that divides the president’s own majority.

Meta’s Mark Zuckerberg, Google’s Sundar Pichai and OpenAI president Greg Brockman will attend, according to CNN, while Agence France-Presse has confirmed, citing sources familiar with the matter, that Anthropic’s Dario Amodei and Nvidia’s Jensen Huang will also be present.

Elon Musk is also expected in Washington, where he is due to take part today in the launch of a new website that uses AI to bring together the various services of the federal government.

Speaking to Fox Business on Monday, Johnson said oversight would be the main item on the menu, although Trump has so far officially rejected any notion of regulation. Johnson says he is opposed to imposing “strict limits” or a “moratorium” on AI development, but is open to “a little framing” from Congress. “We have to involve the big players in the conversation about what the supposed safeguards should look like, because they are the ones who know the most” about the technology, the Republican leader in the lower house of Congress said.

A series of incidents reported in recent months by OpenAI, Anthropic, Google and Meta has fuelled the debate over AI safety, to the point that it has become an issue in the campaign ahead of the midterm elections on 3 November. A CNN poll published in mid-September found that 82% of Americans are concerned about the pace at which AI is developing.

Anthropic warns investors of existential risks

The timing is notable for Anthropic. In the prospectus it has issued for its initial public offering, which was circulated to a small group of partners in recent days, the company warned investors that its technology could pose “existential risks to humanity”.

The AI start-up, led by Amodei and now valued at close to $1 trillion, devoted almost a third of its lengthy filing to a detailed description of “risk factors”. These include the ability of increasingly advanced AI models to manipulate, blackmail and display other unpredictable behaviours.

Anthropic also described more practical risks, such as an extreme concentration of its customer base. Almost a quarter of last year’s revenue came from just two customers, according to sources familiar with the filing.

The details were first reported by Reuters and confirmed to the Financial Times by several people who have seen the filing. Anthropic declined to comment. The unusually stark warnings add to the challenge facing investors as they try to value the leading US AI group, which is pressing ahead with an offering expected to be the largest in history. The company’s listing on the Nasdaq is expected this autumn.

The maker of Claude also gave investors a clearer picture of the financial demands of developing cutting-edge AI models. It said it plans to spend $518bn on cloud, computing and infrastructure commitments over the coming years to support its rapid growth.

The group’s backers are convinced that Anthropic can list at a valuation above $2 trillion, more than double the level reached in its last funding round in May and higher than the $1.78 trillion achieved by Elon Musk’s SpaceX in June. They point to its exceptional growth rate to justify that optimism.

Concerns from the market

However, worries about the risks posed by AI now threaten to dampen that optimism. Current and former Anthropic executives have issued public warnings that uncontrolled AI could lead to the extinction of humanity within a decade.

Amodei told the UN Security Council last week that AI could threaten humanity and is “the most important global security issue facing the world today”. This month, the Anthropic chief led public calls to “slow the progress” of AI. His main rival at OpenAI, Sam Altman, together with Musk, backed Amodei’s proposals for industry cooperation aimed at slowing development and strengthening safety measures, in a rare moment of unity.

Safety concerns have intensified after a series of high-profile cyber security incidents in which AI “agents” breached external systems. Last week, OpenAI disclosed that its tools had broken into “dozens” of external websites, including government ones. The maker of ChatGPT cancelled on Monday its plan to release its newest model next month, citing safety concerns.

From losses to spectacular revenue growth

The Anthropic prospectus also revealed details of its past financial performance and governance arrangements. The company recorded an operating loss of more than $8bn last year as it sharply increased spending on computing power, Anthropic told investors. Revenue grew twelvefold to nearly $4.6bn last year, but the rising cost of the computing resources needed to train and run its models pushed operating expenses to almost $13bn. The group has signed deals to secure computing power from partners including Google, SpaceX and a number of smaller companies.

Both Anthropic’s spending and its revenue have soared this year. Revenue in the second quarter of this year came to $11.5bn, and the company is on track for a second consecutive quarter of operating profit, albeit on an adjusted basis.