Anthropic's Fable Shutdown Accelerates Shift to Open-Source AI and Chinese Models

Anthropic's Abrupt Model Suspension Reshapes AI Market

The US Department of Commerce ordered Anthropic to halt access to its Fable 5 and Mythos 5 models late Friday, citing national security authorities under US export controls on AI 1

1

. The directive prohibited the company from providing these frontier models to anyone outside the US or to "foreign nationals" within the country, including Anthropic's own employees 3

3

. Anthropic responded by suspending access entirely to ensure compliance, leaving customers who depended on these models abruptly cut off.

The timing proved awkward, landing just hours after SpaceX wrapped its record-breaking IPO debut 1

1

. Anthropic had previously positioned its Mythos model as too powerful for unrestricted public release, launching Project Glasswing to provide controlled access to institutions in approximately 15 countries, including US allies like Japan and South Korea 3

3

. The sudden shutdown delivered a stark message: access to closed AI models can vanish at any moment.

Chinese AI Models Surge as Alternatives to Closed AI Models Gain Traction

Investors immediately pivoted toward open-source AI options. Chinese AI models from MiniMax and Zhipu surged on Monday as the market absorbed implications of the Anthropic shutdown 1

1

. Knowledge Atlas, better known as Z.ai, saw shares jump over 30% in Hong Kong trading after releasing GLM-5.2, its latest open-source model 3

3

. The company posted on social media that "frontier intelligence should not belong to only a few people, nor be subject to withdrawal by a handful of rules at any moment," in a clear reference to the Anthropic situation 3

3

.

DeepSeek AI and other Chinese developers now dominate OpenRouter, a popular platform for accessing different AI models. Last week, the top four most-used models came from Chinese companies: DeepSeek, MiniMax, Tencent, and Xiaomi 3

3

. These cost-effective AI models have proven popular not just within China but across developing countries globally, where they represent an attractive balance between price and performance.

AI Vendor Lock-In Risk Drives Companies Toward Sovereign AI

Microsoft CEO Satya Nadella warned about the dangers of dependency despite his company being the principal investor in OpenAI and backing Anthropic with billions of dollars. He wrote Monday that companies need to "build agentic systems that improve over time, while still retaining control over their IP" 1

1

. "The last thing any of us want is a world where every company across every sector is ceding value to a few models that eat everything they see," Nadella added.

Yash Patel, CEO of Applied Compute, which helps companies train and run custom models, said the Anthropic situation "highlighted the significance of owning your own model" 1

1

. He noted that demand for multimodal AI solutions has accelerated dramatically: "What we've been hearing increasingly, probably more so in the last month than the entire year, is the fact that they want a multimodal future. They don't want to be locked into a single vendor."

Paul Triolo, a partner at DGA-Albright Stonebridge Group, observed that this marks "the first time that a government has ordered a model developer to restrict access to a particular model based on nationality" 3

3

. He predicts companies and governments will reconsider their approach to application development and explore sovereign AI options, including non-US models from Mistral, Cohere, and capable Chinese open-source models.

Geopolitical AI Competition Intensifies as China Pursues Different Strategy

China's AI strategy differs fundamentally from America's capital-intensive approach. While US firms optimize for revenue with eyes on trillion-dollar IPOs, Chinese companies face tighter capital constraints and respond by leaning into open-source development, cost efficiency, and global market expansion 4

4

. DeepSeek's parent company claimed it trained its reasoning-focused model for only $6 million, a fraction of OpenAI or Google expenditures 4

4

.

This efficiency stems from structural advantages: AI engineers in China earn approximately 402,000 RMB (about $57,000) annually, far below US salary norms, while China trains one-and-a-half to two times as many AI-relevant PhDs 4

4

. Data centers benefit from cheaper electricity, discounted land, and aggressive local subsidies. Soul Capital's Herry Han describes the dynamic as "like the two sides of tai chi, with its black-and-white yin and yang symbol—each with unique strengths, each pushing the other forward" 4

4

.

Cost Pressures Drive Token-Pocalypse and Model Reassessment

As AI products move toward usage-based pricing, companies face what Patel calls a "token-pocalypse" 1

1

. The era of unlimited token usage has ended, pushing enterprises to route routine work to cheaper models and reserve expensive frontier models for complex tasks. Flo Crivello, CEO of AI agent platform Lindy, switched his company entirely to DeepSeek V4, citing millions in savings and better performance on core use cases while keeping operations on American soil through US hosting providers 2

2

.

Microsoft's exploration of a secured version of China-built DeepSeek V4 to power Copilot Cowork—its agentic assistant and most compute-hungry part of Microsoft 365—demonstrates that even the richest software companies cannot maintain exclusive reliance on closed American models 2

2

. The secured version would keep data on Azure rather than Chinese servers, showing how the open-source question shifts from "if" to "how."

AI researcher Nathan Lambert notes that multiple open-weight competitors, including Moonshot AI's Kimi and Z.ai's GLM models, are now passing thresholds in AI coding tools that Anthropic and OpenAI reached about six months ago 5

5

. GLM-5.2 represents the first open-weight model "that feels right in coding harnesses as a general agent," offering capabilities comparable to recent Claude Code implementations at vastly lower costs 5

5

.

Security Concerns and Technical Soft Power Emerge as Key Issues

The shift toward Chinese open-source models raises security questions that every CEO must address. Many companies already run Chinese AI models somewhere in their technology stack without executive awareness, as vendors and engineers select them based on price 2

2

. Chinese companies are legally required to cooperate with state intelligence, creating potential backdoors and regulatory threats as US-China competition escalates 2

2

.

The situation echoes the costly Huawei ban from telecom networks across the US and its allies, which required billions to rip out and replace infrastructure 2

2

. Not all open models carry these risks—Meta's Llama and Europe's Mistral offer credible alternatives with no ties to Beijing 2

2

.

Neil Shah, vice president of research at Counterpoint Research, notes that China's move validates its broader tech self-sufficiency strategy: "Obviously they're not on the cutting edge because of the export controls, but they have their own silicon and their own software" 3

3

. Chinese firms are building technical soft power through easy access and utility, quietly shaping global AI practices as their models diffuse through startups, universities, and mid-market firms worldwide 4

4

.🟡 disadvantage, but they have their own silicon and their own software" 3

3

. Chinese firms are building technical soft power through easy access and utility, quietly shaping global AI practices as their models diffuse through startups, universities, and mid-market firms worldwide 4

4

.