Anthropic AI pause proposal called Thursday for a globally coordinated slowdown in frontier model development, warning that the most powerful AI systems are showing signs of escaping human control.
The San Francisco-based maker of the Claude AI models said a temporary worldwide slowdown would “likely be a good thing.” It warned that a single-company pause would fail if rivals continued racing ahead.
Anthropic said any workable pause would require major AI companies and governments in several countries, especially the United States and China, to stop simultaneously under verifiable rules.
The company said firms and governments face safety decisions under competitive and geopolitical pressure without a global coordination mechanism. Critics in the industry and White House have accused Anthropic of overstating worst-case risks to slow competitors.
The White House has acknowledged the power of Anthropic’s Mythos model, which remains unavailable to the public because of its cybersecurity capabilities and is deployed only to vetted organisations.
The proposal faces resistance in Washington and Silicon Valley, where officials and executives argue that slowing AI development could give China a strategic edge. President Donald Trump said he discussed AI safety cooperation with China during a recent visit to Beijing.
Read: Anthropic Valuation Hits USD 965 Billion After Funding
Trump also signed an executive order this week giving the government 30 days to conduct a preliminary review of the most powerful US AI models before release.
Anthropic compared the challenge to nuclear arms control but said AI training is harder to detect than missile facilities. The company said AI is already accelerating AI development and could move toward “recursive self-improvement,” while adding that the outcome is not inevitable.
