Amodei Called for an AI Slowdown, but OpenAI’s Astra May Be Changing That

Anthropic is considering releasing a new AI model to counter OpenAI’s momentum following the launch of GPT-6 Astra earlier this month, according to three people familiar with the internal discussions cited by Reuters.

The timing lands awkwardly against CEO Dario Amodei’s long essay urging AI companies to slow the pace of model development, warning that unchecked development could let swarms of AI agents cause potentially hundreds of billions of dollars in damage within a year.

Astra Is Already Eating Into Anthropic’s Enterprise Lead

OpenRouter, a platform that routes developer traffic across different AI models, reported that its users spent more on OpenAI models than on Anthropic’s last week, the first time OpenAI has led that measure in over two and a half years, per Reuters. 

Some investors told Reuters they’re reassessing Anthropic’s position as the leading enterprise AI provider after Astra’s traction, while others argued it poses little near-term threat because of the switching costs in Anthropic’s existing enterprise relationships. 

Anthropic’s own numbers still dwarf its rival’s for now, with OpenAI’s annualized revenue run rate passing $40 billion in July against Anthropic’s $65 billion.

But revenue lead alone hasn’t stopped Anthropic from reportedly treating Astra’s momentum as urgent enough to warrant a faster release than originally planned.

Anthropic Did Take One Real Step Toward Its Own Pledge

To be fair to Anthropic, it hasn’t abandoned Amodei’s framework entirely. 

To put Amodei’s framework into practice, Anthropic selected Accenture as its first “embedded evaluator,” a concrete step toward independent oversight, with both companies committing at least $1 billion over five years toward safety infrastructure. 

That’s a genuine action, not just rhetoric. But accelerating a rival model for market share directly contradicts Amodei’s core point: commercial pressure is what breaks safety restraint first. 

Anthropic’s IPO timeline adds another layer: marketing could now slip past the November midterms entirely, according to people familiar with the plans, even as the company keeps preparing what could be one of the largest IPO in history.

The Company That Sounded the Alarm Is the One Moving Fastest

Notably, Sam Altman backed Amodei’s call for restraint and then delayed OpenAI’s IPO into 2027, citing the safety debate and calling this “an ill-advised moment” to go public.

Anthropic, whose CEO wrote the essay, is reportedly doing nearly the opposite, weighing a faster model release while still pursuing a public offering that could value the company above $2 trillion, surpassing SpaceX’s blockbuster IPO.

Before an IPO, companies have to lay out their biggest risks for investors. It’s tough to brush off a danger as minor when your own CEO just published a massive essay sounding the alarm. 

Anthropic doesn’t need to choose between growth and safety rhetoric forever, but right now, the gap between what its CEO is telling the public and what its business is reportedly doing behind closed doors is the story, not the model itself.

Source: Anthropic considers releasing new AI model ahead of IPO, sources say