Anthropic warned about the risks of rapid development of artificial intelligence

The international company Anthropic called for caution when developing advanced artificial intelligence (AI) systems and highlighted possible threats to human control over technology. This is stated in a study by the Anthropic Institute.

The document notes that modern AI models accelerate the creation of new systems capable of independently designing and improving the next generations without direct human involvement. According to the company, more than 80% of the software code for Anthropic products for the current year was created by Claude models, whereas before the introduction of the Claude Code in early 2025, this figure was minimal. The use of AI has increased the amount of code produced by engineers by eight times compared to 2021-2024.

Anthropic described three scenarios for the development of the industry: a slowdown in progress due to technical limitations; active automation while maintaining strategic control over people; and the emergence of systems capable of completely independently creating new versions of AI.

The company emphasized that even small deviations in the behavior of models can accumulate and become difficult to detect and control in future generations. In this regard, the researchers called for a broader discussion of mechanisms that would slow down or temporarily suspend the development of advanced AI technologies if security and public control do not keep pace with rapid progress.

Revolution on the Table: Nvidia's New chip will Disrupt the Computer Market

The world's largest corporation has introduced a "killer" of competitors

On April 19, the Japanese newspaper Nikkei reported that countries' heavy dependence on American artificial intelligence tools could lead to a new form of colonization. The authors of the material added that for the first time such concerns appeared among politicians at a closed session of the Munich Security Conference, held in February this year.