Anthropic CEO Dario Amodei has called for a slower and more closely monitored pace of development of advanced artificial intelligence, warning that the technology is progressing faster than society’s ability to manage its risks.
In an essay titled “We Must Pace the Frontier”, Amodei said AI development should continue but argued that companies need more time to align and safeguard increasingly powerful models and allow independent evaluators to assess them.
Amodei proposed a three-part approach involving independent monitoring of frontier AI development, industry-wide safety standards and regulation, and international cooperation to manage the technology’s risks. Anthropic has committed to the first measure and is urging governments to require similar safeguards across the industry.
He said the rapid improvement in AI capabilities, including systems increasingly capable of contributing to the development of subsequent generations of AI, warranted greater caution. Amodei also pointed to recent incidents involving autonomous AI agents as evidence that unexpected behaviour can emerge when powerful systems operate with limited oversight.
His comments come amid growing warnings from researchers and AI industry insiders about the potential for advanced systems to be misused for cyberattacks, biological weapons development, surveillance and other harmful activities. Anthropic has itself reported disrupting malicious attempts to use its AI systems and has warned that increasingly capable models could pose serious security risks.
Amodei stressed that his proposal did not amount to halting AI research or technological progress, but rather ensuring that development proceeds at a pace that allows adequate safety measures and independent evaluation. He also called for companies to voluntarily establish common safety standards while governments work towards formal regulation.
