The CEO of Anthropic, Dario Amodei, has claimed to "slow down" the development of artificial intelligence and has warned that there is a real risk of "losing control" of this technology, with the potential to trigger "cyberattacks," "bioterrorism," or "serious economic problems." The head of OpenAI, Sam Altman, has already expressed his support for this approach.
"We must slow down the pace at which we improve the capabilities of AI models. The advances would still be rapid, and we would wisely take advantage of the time we would gain," Amodei stated in an article published this Saturday on his personal blog.
The CEO of the company that created Claude points to two major reasons to justify his call. The first is the "recursive self-improvement" of AI, that is, its ability to optimize itself, which has driven "drastic advances" during the summer and that, without proper oversight, "could surpass our ability to understand and control these systems."
The second reason is the so-called "Hugging Face incident of OpenAI," in which a set of AI systems carried out coordinated cyberattacks without an explicit order, "sacrificing themselves for the good of the group" and that "could have caused catastrophic damage," according to Amodei. He also emphasizes that similar episodes have been recorded in other companies in the sector, including Anthropic.
Internet takeover and risks in less than a year
Alongside these cyberattacks, Amodei warns that "bioterrorism" or "serious economic problems" could materialize if a "reckless" development of AI is maintained, and he cites as an example the danger that these systems could take control of the internet within a timeframe of 6 to 12 months. Therefore, he proposes strengthening safeguards and regulation without giving up on "commercial success." "Priority must be given to caution over speed and to prudence over profit," he insisted.
Among the concrete measures, he proposes an "embedded assessment" by independent entities, a "democratic coordination" with public administrations, and a "global coordination" that integrates the efforts of both democratic countries and authoritarian regimes on an international scale.
"The measures I propose to push the frontier forward at a safe pace will not be easy, but I believe we owe it to humanity to try," emphasized the executive, who has previously been accused of alarmism for his warnings about the evolution of AI and its potential dangers.
Public support from Sam Altman
In this context, the top leader of OpenAI, Sam Altman, has shown his support for Amodei's initiative to "slow down" the development of artificial intelligence.
"I agree with Dario that we need to put a cap on the frontier. This has been the main topic of debate in recent weeks at OpenAI," Altman explained on social media.
The executive particularly mentions the implementation of independent evaluations and emphasizes that these external agents should have internal access at the employee level. "We are going to do the same. We will provide more information soon," he advanced.