Elon Musk warns one AI milestone dwarfs nuclear weapons

17 Sep 2026, 14:33 · The Fresno Bee

Three of the most influential figures in artificial intelligence spent the weekend warning that their technology is advancing too fast. The timing turned a scattered safety debate into a coordinated alarm that rattled global markets within hours.

Elon Musk, chief executive of Tesla (TSLA) and chairman and CEO of SpaceX (SPCX), reshared an April 2023 post on September 12, 2026. In it, he described artificial general intelligence (AGI) as “significantly higher risk than nuclear weapons.”

Musk also warned that no other technology he has observed developing poses the same level of danger, Benzinga reported.

Amodei’s slowdown essay draws rare endorsement from Musk and Altman

Musk’s warning centered on AGI, a theoretical level of machine intelligence that would match or exceed human cognitive ability across every domain, Benzinga noted.

Anthropic CEO Dario Amodei published a roughly 3,900-word essay on the same day, calling on the industry to slow capability development so that safety research has time to keep up.

The essay proposed embedding third-party evaluators inside labs and extending coordination to international limits on AI progress between democratic and authoritarian governments. OpenAI CEO Sam Altman had already signaled concern before the essay appeared.

In a Fortune interview conducted on September 12, 2026, he confirmed that OpenAI would not pursue its initial public offering (IPO) in 2026 as previously expected, describing the current moment as the wrong time to go public while safety and alignment questions remain open.

OpenAI had already filed its S-1 with the Securities and Exchange Commission (SEC) in June 2026.

Musk endorsed Amodei’s essay on X, calling the position correct, and completing a rare alignment among three chief executives whose companies compete directly for market share.

Inside and outside Anthropic, a 10% extinction estimate gains ground

The chief executives’ warnings followed an internal crisis at Anthropic that had become public days earlier, marked by a high-profile departure.

Jacob Coxon, a former pre-training researcher at OpenAI and Anthropic, announced his resignation on X on September 9, 2026 accusing both companies of prioritizing competitive speed toward self-improving systems over adequate safety controls.

More Elon Musk:

Evan Hubinger, alignment science lead and a member of Anthropic’s technical staff, publicly backed Coxon’s assessment and shared a personal risk estimate that carries weight given his direct role in the field.

Hubinger conceded that Anthropic does not yet have a clear plan to solve alignment for superintelligent systems and is not on a path to develop one before those systems arrive.

“We really do earnestly believe AI could kill all humans,” Hubinger wrote on X. “I personally think it is >10% within the next decade.”

Hubinger clarified that the risk does not stem from current models, which Anthropic’s latest internal report described as low-risk.

His concern centers on future systems capable of recursive self-improvement, in which models upgrade their capabilities faster than researchers can build safeguards.

Geoffrey Hinton, the AI pioneer and 2024 Nobel Prize winner in Physics, told BBC Newsnight, on September 9, 2026, that a 10% probability is not unreasonable given that humanity has never created beings that may soon be smarter than itself.

Ramsey Cardy / Getty Images

AI stocks absorb the first coordinated safety shock

Markets delivered a swift verdict on September 14, 2026, with the PHLX semiconductor index tumbling 5.9% in a single session and trimming its 2026 gain to 57%, Reuters reported.

The selloff hit every major exchange from Tokyo to Amsterdam before reaching Wall Street, where futures had already fallen 1.5% ahead of the opening bell.

Citigroup’s equity trading strategy team, led by Stuart Kaiser, warned clients in a September 13 note that a coordinated slowdown in development could squeeze earnings-per-share (EPS) revisions across the technology sector, MarketWatch reported.

The team identified the AI slowdown, midterm election uncertainty, and surging bond yields as converging headwinds that could undermine the EPS story underpinning the broader market rally, Benzinga reported.

Throttling new capability gains would leave near-term revenue from deployed models intact, even as it undercuts the forward growth assumptions embedded in current sector valuations, Kaiser’s Citigroup team noted.

AI-driven earnings growth has powered much of the S&P 500’s 2026 gain, according to Kaiser’s team. They noted that any disruption could ripple beyond chipmakers, affecting cloud and enterprise software valuations.

Why the AI safety push exposes hidden concentration risk

The selloff exposed a deeper structural risk hiding in mainstream portfolios.

Nigel Green, CEO of deVere Group, told InvestorIdeas that investors who deliberately picked technology stocks at least understood their exposure, while those whose holdings drifted toward a single theme through automatic rebalancing never made that active decision.

Somebody who has never bought a tech stock in their life can still be sitting on a concentrated AI position through their pension

Years of strong AI-linked returns have pushed technology concentration into workplace pensions and index funds that were never marketed as bets on a single theme, Green noted.

The key distinction is between slowing new model development and reducing demand for already deployed systems. Green said that the difference will determine whether the selloff signals a repricing or a temporary overreaction.

The AI bet hiding in your retirement account

Green argued that one detail will determine whether this sell-off reprices the sector or passes as a temporary overreaction: whether the slowdown touches only frontier development or also reduces demand for systems already deployed and generating revenue.

Holders of broad index funds and workplace pensions should treat this moment as a prompt to check how much of their portfolio’s 2026 performance traces back to a single AI-driven earnings theme, Green advised. The concentration may not reflect a deliberate choice.

Whether binding slowdown commitments follow the weekend’s public statements remains the open question for anyone holding positions tied to the AI growth narrative, as Green called it.

The Arena Media Brands, LLC THESTREET is a registered trademark of TheStreet, Inc.

This story was originally published September 17, 2026 at 7:33 AM.