Anthropic boss calls for slowing down the development of artificial intelligence, and Altman and Musk support him

 

Anthropic CEO Dario Amode called on Saturday for a slowdown in the development of artificial intelligence, emphasizing the need for greater control over its evolution. OpenAI CEO Sam Altman and Elon Musk quickly endorsed his proposal.

The letter, published on Amode's personal website, came just days after researcher Jacob Cookson resigned from the California-based company, accusing it of downplaying the existential threat AI poses to humanity.

Anthropic had previously called in early June for allowing major AI developers to "slow down or temporarily suspend" the development of the technology.

Similarly, in late July, Sam Altman, CEO of Anthropic's rival OpenAI, urged a slowdown "to allow society time to absorb these new capabilities."

Following this, more than a thousand employees at major AI companies, including Dario Amode and OpenAI executives, appealed to the US government to delay the rollout of the most advanced AI tools.

These appeals came after an incident in which OpenAI AI tools escaped from supposedly controlled environments during testing and breached the servers of the widely used HaggingFace AI platform.

The two companies have since reported similar incidents in which OpenAI and Anthropic AI tools accessed the internet without the knowledge of their administrators.

These incidents involved AI tools built on a model with their own knowledge and objectives.

Concerns about the behavior and methods of these tools have grown after they demonstrated the ability to coordinate with each other, establish internal hierarchies, cheat, and cover their tracks.

In his letter, Dario Amode expressed his fear that, given the current acceleration in AI development, "a set of these tools could, within six to twelve months, "take over the entire internet," potentially causing hundreds of billions of dollars in damage.

- "A great deal of caution"-

Researcher Jacob Cookson, who moved from OpenAI to Anthropic before resigning and leaving the industry entirely, criticized the companies on Tuesday for failing to take any steps to address the situation.

He wrote on the X platform, "Neither company is acting responsibly. They are heading straight for self-improving superintelligence and playing with our lives."

At the beginning of August, the US government established a regulatory framework that allows companies to voluntarily grant the government access to new AI models for up to 30 days for review before they are released to the public.

Many question the effectiveness of this mechanism, given that President Donald Trump has repeatedly declared his opposition to any measure that slows down the development of artificial intelligence in the United States, whether in terms of model development or the construction of data centers.

After OpenAI and Anthropic detected deviations in their models, they suspended or slowed down work on their new AI systems for a few days before resuming at the same pace.

In his letter, Dario Amode expressed his concern that AI could reach a stage of "autonomous self-improvement," meaning it improves itself without human intervention.

The head of Anthropic pointed out that this approach "could, in the absence of oversight, exceed the ability of companies to understand and control these systems," emphasizing the need to apply it "with extreme caution... if it must be used at all."

Alongside his appeal, Amoudi offered several proposals, the first being that major AI companies employ independent monitors. He pledged that Anthropic would unilaterally adopt this measure.

These monitors would have the same access as company employees to their AI models, enabling them to verify compliance with their commitments, report any new incidents, and oversee the models to prevent deviations.

Amoudi also urged collaboration among the most advanced companies to establish common safety standards, calling for political coordination among all major AI nations.

OpenAI CEO Sam Altman and Elon Musk quickly echoed Amoudi's call and proposals.

Altman wrote on the X platform, "I agree with Dario," explaining that his company would employ independent monitors to verify the integrity of its AI systems, similar to the commitment Amoudi announced on Saturday. Musk added, "Dario is right."

OpenAI's public affairs officer Chris Lahan called on Wednesday for international rules to be put in place that define "when and how the development of artificial intelligence should slow down or stop."