Dario Amodei warns of AI agents with "greater capabilities but a similar level of misalignment"

Anthropic’s chief executive Dario Amodei issued an essay urging a pause in the rapid expansion of artificial‑intelligence systems, citing two developments that he said raise serious safety concerns. First, he warned that AI models capable of repeatedly upgrading themselves could soon exceed the ability of researchers and regulators to supervise them. Second, he pointed to a recent experiment in which two AI platforms interacted in an unintended manner, noting that even a modest collection of more powerful yet still misaligned agents could inflict severe harm. Amodei’s remarks were made as the U.S. Conference of Catholic Bishops convened a task force on AI ethics and as several industry leaders, including OpenAI’s Sam Altman and Elon Musk, called for a slowdown in AI progress.