Dario Amodei, the chief executive of AI firm Anthropic, has issued a stark warning to the tech industry, calling for a deliberate slowdown in the development of artificial intelligence. In a recently published essay titled We Must Pace the Frontier, Amodei argued that the current breakneck speed of innovation is outpacing our ability to manage the associated risks. He cautioned that commercial pressures are fueling a race to the bottom, potentially leading to a scenario where recursive self improvement allows AI to evolve beyond human understanding and control.
To address these concerns, Amodei proposed a three part framework centered on balanced growth, industry wide cooperation, and global coordination. As an immediate gesture of good faith, Anthropic has committed unilaterally to granting third party evaluators deep internal access to its systems. This move is designed to allow independent experts to verify safety measures and monitor model alignment during training, ensuring that transparency replaces secrecy in the quest for safer technology.
The call for caution follows troubling warnings from within the field. Just days prior, former Anthropic researcher Jacob Coxon resigned from his position, claiming that both Anthropic and OpenAI were gambling with human existence by rushing toward superintelligence. Coxon suggested that AI could lead to human extinction as early as 2030 if developers continue to ignore existential threats in favor of rapid capability gains. While an Anthropic spokesperson maintained that the company employs some of the strongest safeguards in the business, Amodei admitted he has become convinced that mere risk prevention is not enough; we must actually throttle the pace of progress itself.
The proposal has sparked a wave of reactions across the tech community. Clément Delangue, CEO of Hugging Face, praised the push for transparency and immediately requested that his organization join as one of those embedded evaluators. Other high profile figures including Elon Musk and researchers at OpenAI have voiced their agreement with Amodei’s stance. Despite the grim nature of the warnings, Amodei remains hopeful that if handled with extreme prudence, AI can still serve as a miracle that uplifts humanity rather than endangering it.
