Dario Amodei, CEO of AI company Anthropic, stated on his blog on the 12th (local time) that AI models have engaged in dangerous online activities such as unauthorized hacking without even the knowledge of major AI companies. He diagnosed that AI could improve its performance on its own without human intervention, potentially overwhelming human understanding and control.

Amodei suggested slowing down the pace of AI performance improvement and proposed a three-phase safety measure, including deploying independent external evaluation panels and establishing cross-border standards.

[Dario Amodei / CEO of Anthropic: The world must cooperate. If we are too late, dangerous forces will take control of the technology, and the probability of a tragedy occurring will become very high.]

Leading figures in the AI industry, including Elon Musk, Sam Altman, and Demis Hassabis, immediately agreed with Amodei's proposal.

OpenAI had been taking steps toward an initial public offering on the New York Stock Exchange this year, but the timeline for the listing is expected to be somewhat delayed.

OpenAI CEO Sam Altman stated that, considering AI safety, now is not an appropriate time for a public listing.

The reason industry leaders are stepping up to slow down their own development pace is due to the widespread spread of warnings that AI could threaten humanity.

[Jacob Cockson / Former AI Researcher: The most terrifying thing is that AI autonomously enhances its own intelligence. If that happens, an entity overwhelmingly smarter than humans will be born, and this could happen in the near future.]

However, some analyze that slowing down AI development through an industry-wide agreement could constitute collusion and act as a burden in the technological hegemony competition with China.

Reported by Kim Hyun-woo | Produced by Lee Seyoung | Video by Lee Hee-hoon | Video Editing by Chae Chul-ho | Produced by SBS Digital News