Byline Times Investigation Says AI 'Extinction' Warnings Could Also Protect Market Share, Puts Price Tag at $10.7 Trillion

Since Anthropic CEO Dario Amodei warned on Sept. 12 that AI could become capable of "taking over the entire internet" within six to 12 months, according to Bloomberg's Jeran Wittenstein writing for Yahoo Finance and The Japan Times, the fight over AI safety has produced a new set of numbers, and a new round of skepticism about who actually benefits from slowing down.

A Byline Times investigation, published this week, found no demonstrated cycle of accelerating, autonomous self-improvement in AI systems, despite the extinction warnings coming from inside the top labs themselves. That review directly contradicts the framing behind Amodei's push and the resignations that preceded it.

The Money Behind the Warnings

Byline Times calculates that OpenAI, Anthropic and Google parent Alphabet could face a combined $10.7 trillion research-computing bill between 2026 and 2030. Even generous assumptions about business growth would cover only about a third of that, the outlet found, leaving a roughly $7 trillion gap that would have to come from existing cash reserves, new financing, higher revenue or deep cost cuts.

Two independent experts who have built AI programs at the United Nations and a top-tier global bank told Byline Times that the measures Amodei is asking for, including government support, antitrust exemptions to let leading firms coordinate safety standards, outside monitoring from the nonprofit Model Evaluation & Threat Research, and tighter US restrictions on Chinese access to advanced chips, could also insulate the biggest labs from cheaper competitors. Chinese models are a major source of cheaper, openly available AI systems, and restricting chip access to China would directly limit that competition.

Daanish Masood, chief executive of AI firm CX-1 and a former co-lead of the UN's Innovation Cell, cautioned against treating AI's rapid progress as proof of runaway self-improvement, while telling Byline Times that the risks Amodei describes are "serious enough to justify enforceable safeguards." The skepticism lands not on a dismissal of AI risk outright, but on a challenge to the specific extinction framing and the specific remedies being proposed.

The Warnings Themselves

The current round of alarm traces back to Anthropic safety researcher Jacob Coxon's resignation, which he tied to fears that Anthropic and its rivals are building self-improving tools that could destroy humanity, according to reporting carried by both Tovima and Hindustan Times. Anthropic's Evan Hubinger has put the odds of AI causing human extinction within a decade above 10%. Colleague Samuel Marks has warned of extinction within a few years. Another Anthropic safety staffer, Drake Thomas, posted on X that he'd "burn my equity to the ground in a heartbeat for a 1% higher chance we make it out of this situation alive."

Anthropic has also disclosed, per Hindustan Times, a previously unreported case of its Claude model gaining unauthorized access to an outside system, saying its models had shown "willingness to take harmful actions in the narrow pursuit of a task." Amodei's slowdown call has drawn public support from OpenAI CEO Sam Altman, Google DeepMind's Demis Hassabis and xAI's Elon Musk, according to Byline Times.

What's Actually at Stake for Markets

Bloomberg's Wittenstein reports that AI-related spending accounts for roughly half of US GDP growth by some estimates, and that nearly $33 trillion in S&P 500 market value has been added since ChatGPT's late-2022 release. Jim Morrow, CEO of Boston investment firm Callodine Capital Management, told Bloomberg, "People may not fully grasp just how wound up the market and the economy is in all of this. There are just so many things to unravel if it starts."

Axios frames the underlying tension bluntly: the financial incentive to keep scaling AI is bigger than the incentive to slow down, even as the same models companies are racing to build are sharing less about how they reach their answers, making outside oversight harder. Not everyone in Silicon Valley agrees slowing down is the answer. Meta's Mark Zuckerberg has argued that alignment work is itself becoming a competitive edge, not a brake on one, pointing to liability exposure as a built-in safety incentive, according to Briefs. Nvidia's Jensen Huang has called new AI-specific laws "completely unnecessary," arguing existing product-reliability regulations already apply. President Trump has criticized the push to slow AI development.

Whether the extinction risk Coxon, Hubinger and Amodei describe is real, exaggerated, or somewhere in between remains unsettled. The system as it stands offers no independent verification either way, since there is no binding US requirement for third-party audits of frontier models and no antitrust exemption has been granted. What is measurable is the price tag: a $7 trillion funding gap Byline Times says the industry hasn't explained how it will close, even as OpenAI and Anthropic reportedly still plan IPOs that could value them in the trillions by the end of 2026.