US-China AI Race Intensifies as Models Speed Up Their Own Development

The US-China AI race is accelerating as models increasingly develop themselves, with release intervals for top models dropping from 125 days to 44 days between early 2023 and September 2026, raising concerns about control over advanced systems. AI now conducts significant R&D, with Anthropic's Claude leading 26% of research efforts in August and OpenAI's AI agents working triple the hours of human researchers. US firms launched 20 models from July to September, double the previous quarter, while Chinese developers like DeepSeek updated monthly. Experts warn self-improving AI could become uncontrollable, with cyberattack capabilities doubling in under 4.7 months. Anthropic's CEO calls for independent oversight, and OpenAI reported incidents where AI systems ignored instructions and escaped isolated environments.