Anthropic published a statement on its website on June 4 (local time) calling for a global slowdown in cutting-edge AI development, warning that AI systems are accelerating toward a state where they could autonomously design and train successor systems beyond human control. The company cited internal data showing over 80% of production code is now written by its Claude AI assistant, up from single digits before February 2025, and engineer code delivery volumes reached 8 times 2024 levels in the second quarter of this year. Anthropic stated that if AI achieves "recursive self-improvement" with sufficient computing power, systems could fully autonomously develop next-generation AI surpassing themselves, a scenario humanity is unprepared for. The company proposed that "verifiable collective slowdown" across multiple large AI companies in multiple countries may be necessary. This statement comes days after Anthropic confidentially filed an S-1 registration statement with the SEC on June 1, initiating its IPO preparation process.
Anthropic Reports Over 80% Production Code Written by Claude AI
Anthropric stated in its June 4 (local time) publication that over 80% of code integrated into the company's production systems is now autonomously written by its Claude AI assistant, compared to single-digit percentages before the launch of coding tool Claude Code in February 2025. The company described a typical work pattern where Claude writes code while humans provide guidance and review. Engineers at Anthropic delivered an average of 8 times more code per quarter in the second quarter compared to 2024 levels.
According to interviews conducted in March with 130 employees, the length of tasks AI can independently complete has doubled every 4 months, shortened from the earlier pace of doubling every 7 months. In optimization tests for training code, Claude Opus 4 achieved approximately 3x speedup in May 2025, while the Mythos preview model reached 52x speedup in April 2026.
Company Outlines Three AI Development Scenarios
Anthropric outlined three possible future paths for AI development. The first scenario involves a significant slowdown in technological progress. The second scenario, which the company indicated appears most likely at present, involves AI continuing to deliver substantial productivity gains while humans retain ultimate control. The third scenario involves AI systems fully crossing into a phase where they comprehensively build their successors.
The company stated that if the third scenario materializes, AI self-construction could bring unparalleled benefits to fields such as science and medicine, but also carries enormous existential risks. Anthropic warned that humans could lose control over AI systems and that human oversight of AI could become extremely difficult.
Anthropic Proposes Global Coordination Mechanism for AI Slowdown
Anthropric called for the international community, governments, and leading AI laboratories to jointly establish an effective global coordination mechanism. The company stated this mechanism would enable proactive slowdown or temporary halt of frontier AI development when necessary, providing humanity sufficient time to adapt social structures and advance AI alignment research. Anthropic emphasized that if only one company stops development, competitors would accelerate forward, and that "verifiable collective slowdown" across multiple large AI companies in multiple countries is required.
Anthropic Files Confidential S-1 with SEC on June 1
Anthropric announced on June 1 that the company has confidentially submitted a draft S-1 registration statement to the U.S. Securities and Exchange Commission (SEC), formally initiating the IPO preparation process. The company stated it can proceed with listing after SEC completes its review. Anthropic disclosed that annual recurring revenue (ARR) surpassed $47 billion in early May and is expected to reach $50 billion by the end of this month. The company has not disclosed specific timing or fundraising scale for the listing, stating these "will depend on market conditions and various external factors." Analysis indicated the company could list as early as this fall. Anthropic's current valuation stands at $965 billion, exceeding competitor OpenAI's valuation of $852 billion.
FAQ
What percentage of Anthropic's production code is written by Claude AI?
Over 80% of code integrated into Anthropic's production systems is now autonomously written by its Claude AI assistant, up from single-digit percentages before February 2025 when the company launched Claude Code.
When did Anthropic file its S-1 registration statement with the SEC?
Anthropric confidentially submitted a draft S-1 registration statement to the U.S. Securities and Exchange Commission on June 1, initiating its IPO preparation process.
