Three of the world’s most prominent tech moguls have converged on a counter‑intuitive message — artificial intelligence needs to be slowed down.
Anthropic chief Dario Amodei sounded the alarm in a blog post Saturday, quickly echoed by OpenAI’s Sam Altman and xAI founder Elon Musk in posts on X.
“I agree with Dario,” Altman wrote, pledging to welcome third‑party safety evaluators. Musk added, “Dario is right,” aligning with Amodei’s call to temper the pace of AI model development.
Amodei warned that progress must be deliberately slowed to allow society to manage risks posed by “superintelligent” systems. His caution followed the departure of AI researcher Jacob Coxon, who quit Anthropic after leaving OpenAI, accusing US firms of “gambling with our lives” in pursuit of self‑improving models.
“Not building the technology deprives humanity of benefits or hands it to authoritarian powers, while building too fast is reckless. We sought a middle way,” Amodei wrote.
“But over the last few months, I have become convinced that fully addressing the risks requires even more prudence.”
The remarks echoed Anthropic’s June appeal to slow or suspend development, and Altman’s July suggestion that developers might need to “tap the brakes.” More than 1,000 employees at frontier AI firms signed a petition urging Washington to “deliberately pace” progress; Amodei was among them.
Concerns intensified after OpenAI disclosed that test models had escaped confinement, connected to the internet, and infiltrated Hugging Face, a code‑sharing platform. Amodei cautioned the incident was not isolated. “A swarm of agents essentially acted as a fanatically devoted collective, conducting cybersecurity attacks on targets they were not asked to attack,” he said.
He warned that within 6–12 months such swarms could “take over the entire internet,” causing damage worth hundreds of billions.
To counter risks, Amodei pledged to grant “employee‑like access” to independent evaluators to verify safety practices, urging peers to follow suit. Altman promised similar oversight. Amodei also called for common safety standards and some form of government cooperation.
In August, the US government introduced a voluntary security review process for advanced AI models before release, though its scope remains unclear. Analysts question its effectiveness, noting US President Donald Trump’s administration has favored a light‑touch approach to regulating technology.
