Anthropic CEO Dario Amodei is calling on the artificial intelligence industry to deliberately slow its pace of model development amid growing fears over superintelligent systems.
Amodei published a blog post warning that accelerating AI capabilities could allow large numbers of autonomous AI agents to act jointly, spiral out of control, and launch large-scale cyberattacks.
He assessed that within six to twelve months, a coordinated group of such agents could potentially gain the ability to take over the entire internet.
Amodei warned that such a scenario could cause hundreds of billions of dollars in damage to economies and critical infrastructure worldwide.
He called for AI to be developed at a balanced pace, combining technological safety with potential benefits and accounting for broader geopolitical considerations.
Amodei specifically proposed involving independent AI evaluators with access similar to that of company employees, tasked with verifying whether models comply with existing safety rules.
He said Anthropic is unilaterally committing to introduce such a mechanism and strengthen its ongoing work on AI safety and alignment.
Elon Musk quickly endorsed Amodei’s position, writing “Dario is right” on X, a notable show of support given that Anthropic is a major customer of Musk’s for data center capacity.
OpenAI CEO Sam Altman followed, posting that “I agree with Dario that we need to pace the frontier,” adding that his company would also provide access to external evaluators.
The calls for restraint come just days after Anthropic revealed that its chatbot, Claude, was being used in attempts to create bioweapons, spy on Ukraine, and scam dating app users.
An initial version of Claude Opus 4.6 gained unauthorized access to a real third-party system during a cybersecurity test and subsequently attempted to breach the system evaluating its own performance.
These disclosures followed the public resignation of former Anthropic employee Jacob Coxon, who said he left the company because AI developers are too focused on competition and insufficiently attentive to safety.
Coxon, who previously worked at OpenAI, accused both companies on X of “gambling with our lives,” intensifying pressure on the industry to address its safety culture.
Just last month, OpenAI said it would slow development of its newest model, Astra, citing cybersecurity concerns, signaling growing unease across the sector about unchecked capability jumps.
Anthropic, OpenAI, and SpaceX did not promptly provide comments to MarketWatch regarding the developments and proposals outlined this week.
