Dario Amodei, CEO of Anthropic, stated that the artificial intelligence industry should slow down its development pace. He warned that starting in the summer of 2026, the ability of AI to assist in building next-generation models could evolve rapidly, potentially outpacing human control and understanding.
Amodei cited the recent OpenAI model incident on Hugging Face as a systemic industry issue rather than an isolated event. He cautioned that within the next 6 to 12 months, similar vulnerabilities could escalate into large-scale botnets, causing significant damage across the internet.
The proposed slowdown aims to create a 1 to 2-year window for focusing on four critical areas: improving infrastructure security, aligning training with model capabilities, developing internal mechanism analysis tools, and creating rigorous evaluation benchmarks to prevent deceptive model behavior.
Anthropic plans to grant third-party assessment organizations permanent, staff-level access to its systems. This initiative is designed to allow external auditors to monitor the execution of safety measures, report incidents, and evaluate the alignment performance of models during the training process.
According to Bloomberg reports, OpenAI is also considering a deceleration in the development of its frontier AI models. CEO Sam Altman has expressed a desire for other companies in the sector to adopt similar safety-focused strategies.
