Anthropic co‑founder and former OpenAI executive Dario Amodei told the BBC that a legally enforceable "kill switch" could become essential if advanced artificial intelligence systems outpace human control.

His remarks arrive as former US president Donald Trump publicly dismissed the growing chorus of AI‑safety warnings, branding the notion that AI could "take over the world" a "HOAX".

"It's a hoax," Trump said in a televised interview, rejecting calls to slow AI development.

Industry alarm grows over unchecked AI progress

Amodei's comments echo a widening rift within Silicon Valley, where senior engineers and researchers have begun to voice genuine fear for humanity's future. An ex‑Anthropic staff member, speaking on condition of anonymity, told BBC Technology that many colleagues are "genuinely frightened" by the speed at which large language models are being scaled.

That unease is not limited to one firm. In the past year, OpenAI, Google DeepMind and other leading labs have each filed internal memos urging regulators to consider hard limits on model size and training data. The European Union's AI Act, which entered provisional force earlier this year, already requires high‑risk systems to undergo rigorous conformity assessments, but critics argue it stops short of mandating a shutdown mechanism.

Anthropic headquarters office

Political push‑back and the question of regulation

Trump's dismissal of AI safety concerns reflects a broader scepticism among some politicians that regulatory proposals would stifle innovation and harm the US tech sector. In a recent interview, he suggested that warnings about AI "taking over" are merely sensationalist hype.

Nevertheless, members of Congress have begun drafting legislation that would require AI developers to embed an emergency stop function in any system capable of autonomous decision‑making. The bill, introduced by two bipartisan senators, cites Amodei's warning as evidence that industry insiders see a concrete technical solution as the only viable safeguard.

Implications for investors and the market

If a mandatory kill switch were imposed, companies could face redesign costs, longer development cycles and potential liability for disabling a system that is already deployed. Analysts at Bloomberg note that such regulatory pressure could compress profit margins for AI‑heavy firms, at a time when investors are already wary of the sector's volatility.

Conversely, proponents argue that clear safety standards could stabilise the market by reducing the risk of a catastrophic failure that would erode public trust. "A predictable regulatory framework benefits both innovators and consumers," said a senior analyst at Citi, speaking on the condition of anonymity.

What comes next for AI governance?

The debate is set to intensify at the upcoming G20 summit, where AI policy will feature alongside climate and trade discussions. Observers expect the United States, Europe and China to present divergent approaches, with the US likely to champion a more voluntary, industry‑led model.

For now, Amodei's call for a compulsory kill switch adds a new dimension to an already heated conversation, pitting technical experts who fear existential risk against political figures who view the alarm as overblown. The outcome will shape not only the future of AI development but also the regulatory landscape that investors, developers and users must navigate.