Washington's decision to restrict foreign access to Anthropic's flagship models backfired within days. The export control, intended to preserve America's AI advantage, instead accelerated development of competing systems globally.
Cohere reports a surge in government inquiries following the ban. DeepSeek closed a $7.4 billion funding round, the largest in its history. Chinese AI labs responded by slashing token prices up to 99 percent, making their models dramatically cheaper than American alternatives.
The restriction covers Anthropic's Claude models, which the company positioned as safer, more capable alternatives to other frontier systems. Rather than consolidate American dominance, the policy created space for rivals to capture customers seeking advanced AI without export restrictions.
DeepSeek's financing comes as China accelerates its AI race independent of Western technology. The company, backed by Chinese institutional investors, now operates with substantially more capital to scale infrastructure and research. Competing Chinese labs rapidly undercut U.S. pricing, making switching costs negligible for price-sensitive customers.
Cohere, a Canadian AI company, gained unexpected leverage. Government agencies and enterprises now view Cohere as an accessible alternative to restricted American models. The company did not disclose funding amounts but characterized the inbound interest as exceptional.
The policy exposed a blind spot in export control strategy. Restricting one company's access does not eliminate demand for frontier AI capabilities. It redirects that demand toward competitors operating in less regulated jurisdictions. China's labs gain customers, funding, and real-world data without competing directly against American restrictions.
Separately, researchers discovered 144 poisoned npm packages targeting developers working with AI systems. The malicious packages contained credential-stealing code, transforming the open-source supply chain into an attack vector. The discovery highlights how AI development pipelines depend on unvetted dependencies, creating security exposure at scale.
The week illustrates compounding risks