Amodei has urged AI companies across the globe to unanimously pause or slow down development until the social implications of the technology can be fully gauged. Screengrab: ANI
'AI risks are here, ' says Anthropic CEO as he pushes for tougher oversight
Anthropic CEO Dario Amodei urges mandatory AI testing, stronger regulation and powers to block high-risk AI systems before deployment
Expressing concern over the rapid development of Artificial Intelligence (AI) and its potential adverse impact on society, Anthropic co-founder and CEO Dario Amodei has said that he was apprehensive that something would “go wrong”.
"I worry that something will go wrong," said Amodei in a Bloomberg documentary, adding, "Are we doing literally everything we can? We're certainly trying our best.”
Anxiety over AI’s impact
The remarks come at a time when a Reuters poll has revealed that as much as 50 per cent of Americans are apprehensive that the increasing development of AI could put them or their family members’ jobs at risk. The survey shows that the fear of loss of employment is spread evenly across age, gender and education level among the respondents.
Also Read: Anthropic warns AI could one day build its own successors, challenging human control
Amodei, who has been vocal about approaching rapid advancement of AI with caution, has even urged AI companies across the globe to unanimously pause or slow down development until the social implications of the technology can be fully gauged. However, he is doubtful whether such an understanding can actually be reached.
Push for tighter oversight
He also argued that there is a need for stricter government regulation on the development of AI systems, adding that such systems must be subjected to safety checks before being released.
Also Read: Claude concerns: Why and how Anthropic wants to pause AI development
As to whether there was any possibility of AI being banned, Amodei said that it was “possible”, adding "The main way I could see AI being banned or blocked is if something really went wrong. And if something really went wrong, then maybe it deserves to be.”
He has also argued, in a policy essay published on Wednesday, that governments should have the authority to block the release of AI systems if they are found to be posing unacceptable risks when subjected to independent evaluations.
Call for mandatory testing
Amodei further stated that voluntary transparency from AI companies was no longer adequate as the risks are clearly here."
He said that advanced AI models should be subjected to compulsory independent assessments before they are made publicly available, particularly to evaluate risks related to cybersecurity, biological threats and the possibility of systems operating beyond human oversight.
Also Read: Olah and the Pope: What's Anthropic doing in the Vatican?
According to him, authorities should have the ability to delay or prevent the deployment of models that do not meet established safety standards.
"In addition to transparency, I now believe frontier models should face mandatory third-party testing for cyber, bio, and autonomy risks - with the power to block or revoke deployment of models that pose catastrophic risk," Amodei separately said in a post on X.