Asian AI startups pivot to local models amid Anthropic export bans

Sakana AI’s new model, Fugu, debuted in Tokyo this week, explicitly marketed as a hedge against the sudden loss of American AI access. Co-founder David Ha argues that relying on a single foreign provider for national infrastructure now carries an unacceptable risk, positioning Fugu as an orchestration layer capable of coordinating multiple models. Despite the timing, the startup insists its launch was coincidental, rooted in research presented at ICLR this spring, and maintains that U.S. technology remains vital for Asian markets.

In Beijing, the response has been more direct. Cybersecurity firm 360 unveiled Tulongfeng and Yitianzhen, tools designed to automate vulnerability discovery and incident response. Founder Zhou Hongyi framed these releases as a strategic necessity, warning against a "one-way transparency" where advanced cyber-defense capabilities are hoarded by select global actors. While Sakana seeks to preserve collaborative access, 360’s move signals a shift toward sovereign AI independence. For Anthropic, which hit a $47 billion run-rate revenue in May 2026, the long-term impact of these regional alternatives remains uncertain, though the gap is clearly closing.