Restriction as accelerant
TechCrunch captured the strategic cost of the Anthropic export ban in one dispatch from its third week: Asian companies simply shipped alternatives. In Beijing, cybersecurity firm 360 used the ISC AI 2026 conference to unveil Tulongfeng, an AI tool for automated software-vulnerability discovery it claims can go head-to-head with Mythos, alongside Yitianzhen for automated cyber defence and incident response. Founder Zhou Hongyi framed the launch in explicitly geopolitical terms - vulnerability-finding AI as a national strategic asset - and warned of one-way transparency, where some countries can probe software for weaknesses while others cannot. In Tokyo, Sakana AI launched Fugu: a compact orchestrator that routes work across a swappable pool of external models, which Sakana says stands shoulder-to-shoulder with Fable 5 and Mythos Preview. A spokesperson called the timing entirely coincidental - the underlying research predates the ban and appeared at ICLR this spring - but the company's site advertised frontier capability without the risk of export controls, and co-founder David Ha argued the ban made single-provider dependence impossible to ignore.
The numbers and the diplomacy
The commercial exposure is not small: Anthropic's run-rate revenue crossed $47 billion in May 2026, with its dependence on Asian enterprise customers undisclosed. The policy argument went global in parallel - Sakana's Ren Ito attended the G7 summit in Evian and argued publicly that Washington should preserve access for close allies rather than treat frontier models as export-controlled weapons, while India debated a $5 billion sovereign AI fund. Even with Fable and Mythos access restored on July 1, the core observation stands: local alternatives trained for local language and nuance now exist, funded and launched, marketing on a promise Washington wrote for them.
The strategic read for buyers and vendors
- For enterprise buyers in Asia and Europe, the market handed you options and leverage: regional models with data-residency and no-export-risk guarantees are credible enough to pilot as a secondary lane. Use them to price your primary US contracts even if you never switch.
- For US vendors and their partners, the lesson is churn math: two weeks of unavailability created permanent competitors. If you resell or build on American frontier models abroad, continuity provisions - escrow, named fallback models, service credits for regulatory outage - are now a sales requirement rather than a nicety.
- Watch the security-AI category specifically: 360's positioning confirms offensive-capability models are treated as sovereign assets on both sides of the Pacific. Expect procurement restrictions, certification regimes, and allied-access frameworks to form here first - plan any security-AI roadmap with jurisdiction in mind.