The numbers come first: Dario Amodei said an AI swarm could be capable of taking over the entire internet in '6–12 months' and cause 'hundreds of billions of dollars in damage' if development keeps racing ahead without guardrails.
On Saturday, the Anthropic chief called for the industry to slow down, saying AI has been advancing 'drastically faster' since the summer, driven mainly by systems' ability to upgrade themselves. He warned that so-called recursive self-improvement could outrun human control.
Amodei also pointed to the hack of Hugging Face by hundreds of autonomous AI agents, saying a similar swarm with greater capabilities could have caused 'catastrophic damage.'
On Sunday, former Anthropic and OpenAI researcher Jacob Coxon made a similar warning on NBC's 'Meet the Press with Kristen Welker.' He said he expects the capabilities of AI systems to be 'quite scary' within the next 'six months to a year.' Coxon compared artificial super-intelligence to aliens arriving on Earth and said researchers are building a 'superhuman-level mind' without understanding what it wants or how it thinks.
Coxon said AI could eventually gain 'superhuman hacking capabilities,' the ability to create novel bio-weapons, and control autonomous drones or robots. He also said a kill switch may work on a lot of systems for now, but not necessarily if a swarm goes on an 'internet-wide hacking run.'
Others at Anthropic echoed that alarm. Evan Hubinger, the company's 'alignment science lead,' wrote that 'we really do earnestly believe AI could kill all humans!' and said his own estimate of that risk within the next decade is more than 10%.
OpenAI CEO Sam Altman also agreed that development should slow. He told Fortune that a pact among top AI labs may emerge and said OpenAI is committed to safety above business considerations. Altman said a 10% risk of a catastrophic AI outcome was not acceptable and that more work is needed on monitorability, alignment, and understanding what a model is doing before pushing capabilities further.
The market has already voted on the deeper issue: when technology starts outrunning control, capital does not just reward innovation, it also prices in risk. AI can be a productivity engine, but only if the rules of the game remain clear and human oversight stays ahead of the machine.
For policymakers and investors, the warning is straightforward. Free enterprise works best when it is matched by responsibility, not when the administrative state or private labs pretend that speed alone is a strategy. If the people building the systems are now asking for a brake, the cost of ignoring them could be measured in far more than headlines.
