Anthropic Urges a Global Freeze on AI Training Due to Self-Improvement Risks

Anthropic Calls for a Slowdown in Frontier AI Development

According to Главком: The company behind the Claude language model, Anthropic, is pushing for international talks on slowing down or temporarily halting frontier AI training. The main concern is the threat of recursive self-improvement in artificial intelligence. According to data released as of June 5, 2026, over 80% of the code in Anthropic's production base is now generated by Claude itself, highlighting how deeply this technology already shapes development workflows.

AI Productivity Gains and Regulatory Challenges

In the second quarter of 2026, Anthropic engineers can integrate eight times more code per day compared to 2024, signaling a dramatic shift in productivity driven by cutting-edge AI tools. To address the risks, Anthropic proposes a strict international verification mechanism to enforce a training pause, warning that without global oversight, a moratorium could advantage those who continue development in secret.

The regulation of AI is compared to controls on nuclear or missile weapons, yet training digital models is far easier to conceal than building military factories. Critics, however, see the initiative as a move to slow down smaller competitors, potentially freezing the current market power balance. Many experts worry that moratoriums could entrench existing players.

Anthropic plans a series of roundtables with policymakers, academics, and civil society representatives to debate these issues. The conversation around AI regulation could reshape the market, including access for Ukrainian developers. According to Cloudflare data, automated bots have for the first time surpassed humans in global internet traffic volume, underscoring the urgency of these discussions.

Anthropic's call to decelerate frontier AI development reflects mounting safety concerns that could have major consequences for the global tech ecosystem. International dialogue on AI regulation may prove a crucial step toward control mechanisms that prevent potential threats from recursive self-improvement. At the same time, it could influence the competitiveness of various companies, particularly regarding technology access for Ukrainian developers.

The call for a slowdown in AI development comes amid growing concerns about the risks associated with advanced technologies. This is not the first time experts have cautioned about the potential loss of control over artificial intelligence. A recent statement from a Nobel laureate highlights the urgency of these discussions, emphasizing the critical need for regulatory frameworks to ensure humanity retains oversight over AI advancements.