Anthropic CEO Dario Amodei has unexpectedly called for a global deceleration in training advanced artificial intelligence models. This warning from a top industry pioneer highlights escalating safety concerns regarding catastrophic global risks.
$35 billion
Anthropic valuation
⏳ Time Machine
How today’s news fits into the bigger picture
1975
Asilomar conference pauses genetic research
Leading geneticists voluntarily halt specific recombinant DNA experiments to draft strict safety guidelines, creating a historical precedent for self-regulation in high-risk scientific fields.
2021
Anthropic is founded by safety defectors
A group of senior researchers leaves OpenAI over concerns that the company is prioritizing commercialization over safety, establishing Anthropic to focus on building steerable, safe AI systems.
2023
Industry leaders sign the pause letter
Thousands of technology experts and executives sign an open letter calling for a six-month pause on training models more powerful than GPT-4, though major commercial labs continue their scaling race uninterrupted.
Today
Anthropic CEO Dario Amodei publicly calls for a coordinated slowdown in training frontier artificial intelligence models.
What happens next?
International tech regulators plan to convene a safety summit in late 2026 to propose unified testing standards for advanced AI.
In a surprising development on Saturday, Anthropic CEO Dario Amodei publicly urged competitors and governments to slow down the breakneck pace of training frontier artificial intelligence systems. Anthropic, the multi-billion-dollar creator of the Claude AI assistant, was founded on a commitment to safety, but this direct call for an active pause marks a major shift in executive rhetoric. Amodei warned that models trained over the next **18 months** could possess dangerous capabilities, particularly in aiding cyberattacks or assisting in biological warfare. While tech giants continue to pour billions into larger neural networks, this unprecedented warning from a sitting industry leader introduces severe regulatory risks and highlights the growing internal panic that the capabilities of these digital minds are rapidly outstripping our ability to control them.
💭 If you're wondering…
Advanced AI models can quickly synthesize vast amounts of biological data, which could allow bad actors without scientific training to design dangerous pathogens.
Did this story help?
Official sources
Knowledge Chain — tap a concept