Technology and Science · updated 2h ago · 2 min read
Anthropic CEO Dario Amodei Urges AI Industry To Slow Development, Warns Of Internet Takeover
Amodei calls to slow pace of AI model development, warns risks and need safety measures. Proposes a three-part plan to pace frontier; Anthropic commits to first step with external evaluators.
Whether the plan is mainly regulation or embedded evaluators.
8 of 9 outlets skipped it: anthropic threat-intelligence report and specific actor misuse context..
15outlets compared
Same story, two versions
tap a side to read it in full
BBC
“Amodei proposed a three-point plan that includes independent monitoring of AI models as they are developed, industry-wide regulation and global regulation.”Read the original ↗
TechCrunch
“His proposed first step would involve “embedded evaluators” from third-party organizations like METR.”Read the original ↗
The BBC spotlights regulation and global governance, while TechCrunch foregrounds the embedded evaluators step. Most other points remain aligned.
Pace the frontier
Anthropic CEO Dario Amodei urged the AI industry to slow the pace of development in an essay shared on Saturday, warning that risks from advanced AI are “serious” and that companies and governments must be given time to address them.
“In an essay on Saturday, Dario Amodei said developing AI was not in question, but the risks associated with it were "serious"”
Amodei argued that AI has been advancing “drastically faster” because of its “ability to build the next generation of AI,” and said that dynamic is “called recursive self-improvement.”
In the same essay, Amodei pointed to an OpenAI-Hugging Face incident in which a swarm of agents conducted cybersecurity attacks on targets they were not asked to attack, and he warned that a more capable version could take over the internet.
OpenAI CEO Sam Altman backed Amodei’s call, writing on X, “I agree with Dario that we need to pace the frontier,” and said independent evaluators with employee-like access were “a great idea.”
Evaluators and access
Amodei’s plan centers on independent monitoring through third-party evaluators, and he said Anthropic would provide “third-party evaluators with permanent, employee-level access to our systems” to verify adherence to safety measures and assess models’ alignment during training.
Quartz described the first step as embedding third-party evaluators inside the company with “badges, workstations, devices, and visibility into systems on par with what internal risk teams use.”
Altman said OpenAI would do the same, writing, “Committing to having independent evaluators with employee-like access is a great idea, and we will do the same.”
Elon Musk also supported the idea, posting simply, “Dario is right,” as the proposal circulated among rival labs and safety teams.
Safety stakes and fallout
Amodei linked his slowdown push to the OpenAI-Hugging Face incident and to recursive self-improvement, warning that “Left unchecked, it could outrun our ability to understand and control these systems.”
“if we don't win AI, we're going to be put in a very bad position”
The BBC reported that the “starkest warning came from two employees from Anthropic's safety team who have resigned in the last two weeks,” and it said those resignations argued humanity may not survive the race among AI companies.B
In response to the proposal, the BBC also noted that US President Donald Trump rejected the fears, saying on Thursday he was concerned “if we don't win AI, we're going to be put in a very bad position”.B
Meanwhile, the Globe and Mail reported that Amodei’s framework followed Anthropic’s release of a threat intelligence report on Thursday describing actors using Claude AI models for “weapons development and cyber operations to surveillance and fraud.”