Anthropic CEO calls for a slowdown in AI development due to safety risks

Dario Amodei, CEO of AI company Anthropic, has publicly called for a controlled deceleration in the development of artificial intelligence systems to address critical safety concerns. Amodei warned that without a temporary slowdown, powerful AI models could, within six to twelve months, reach a stage where autonomous "agent swarms" might create persistent botnets and seize control of internet infrastructure, potentially causing hundreds of billions of dollars in damage. He emphasized that progress would continue but that the additional time is necessary to ensure safety mechanisms can keep pace with rapid technological advancements.

The proposal received rare public support from major industry leaders, including Elon Musk and OpenAI CEO Sam Altman, who confirmed that AI safety has been a central topic of internal discussions at OpenAI. Amodei, who left OpenAI in 2020 due to differences regarding safety standards, noted that uncontrolled evolution of AI could facilitate cyberattacks and biosecurity threats. These statements coincide with the departure of researcher Jacob Coxon from OpenAI and an annual Silicon Valley investor gathering in San Francisco. While some observers suggest these public warnings may be a strategy to influence investor sentiment, the discourse highlights a growing consensus among developers regarding the existential risks posed by advanced AI.