TL;DR
The United States imposed and later lifted controls on two deployed Anthropic models while introducing government review for other frontier systems. Reuters also reported that China’s Commerce Ministry discussed restrictions on overseas access to models, raising questions about whether future Chinese frontier releases will remain open.
The United States and China have both taken steps toward tighter control of frontier artificial intelligence models, challenging the assumption that European organizations will retain steady access to advanced foreign systems. US authorities gated deployed and forthcoming models in June, while Reuters reported on July 7 that China’s Commerce Ministry had discussed possible limits on overseas model access with major technology companies.
Washington’s actions made model access controls an operating policy tool. On June 2, 2026, Executive Order 14409 established classified capability benchmarks, an NSA designation process for covered frontier models and a voluntary 30-day pre-release government review window, with provisions set to harden on August 1, according to the source material.
On June 12 and 13, the Commerce Department imposed export controls on Anthropic’s Claude Fable 5 and Mythos 5, restricting access by foreign nationals worldwide. The reported trigger involved a jailbreak method, but its severity was not independently confirmed. Anthropic disputed the proportionality of recalling widely deployed products over what it described as a narrow finding. The government later lifted the controls after a weeks-long dispute, restoring global access.
OpenAI then previewed GPT-5.6 Sol on June 26 through a staggered process requiring customer-by-customer government approval, initially covering about 20 organizations. In China, Reuters reported that the Ministry of Commerce, known as MOFCOM, had spoken with Alibaba, ByteDance and Z.ai about possible overseas-access restrictions involving unreleased and open-weight models. The report described discussions, not an issued decree.
The China Open-Weight Window
Both Superpowers Just Put Their Hands on the Doors
The load-bearing assumption under Europe’s local-first economics is being stress-tested — on both sides, in the same month
Two doors, one month
The American door: gating became a regime
- Jun 2: EO 14409 — classified benchmarks, 30-day pre-release window
- Jun 12–13: export controls on two deployed Anthropic frontier models — trigger reported, not independently confirmed; company disputed; later lifted
- Jun 26: GPT-5.6 Sol ships behind customer-by-customer government approval
- The temporariness taught its own lesson about US supply reliability (CEPA)
The Chinese door: hinges of a subtler design
- May: Supreme People’s Court journal roundtable on tiered open-source governance
- Jun: Manus acquisition unwound; sweeping cross-border investment rules
- Jul 7: MOFCOM talks reported — incl. unreleased and open-weight models; discussions, not decree
- Tiering already visible: Qwen 3.6 open, Qwen 3.7 Max API-only
THE STRUCTURAL ASYMMETRY
The US can gate its closed models; it cannot gate published weights. Every gated American model makes the ungatable open alternative relatively more attractive — a feedback loop that is now official-policy-shaped. Beijing’s version inverts it: keep the mid-tier open for soft power, move the frontier behind the counter.
Five moves while the width is known
The verdict: existing checkpoints are safe — the refresh cycle isn’t. The practical question for 2027 is not “will GLM-5.2 vanish?” but “will GLM-6 ship open at launch?” Watch launch mode, not launch benchmarks. The window’s width will be announced in a launch post, not a policy paper.
AI frontier model access control software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Europe Faces a Model-Supply Risk
The developments matter because European companies and public bodies increasingly base local, self-hosted AI systems on model weights released by foreign laboratories. Four of the five leading open-weight model families cited in the source material are Chinese, making continued access to new releases a material part of the economic case for European AI sovereignty.
Published weights generally cannot be withdrawn from organizations that have already downloaded and lawfully mirrored them. Governments can, however, restrict future releases, updates and access to closed services. The immediate risk is not that existing checkpoints disappear; it is that the regular arrival of stronger replacements slows or ends. That would affect performance planning, infrastructure purchases and the cost of maintaining competitive systems.
The two countries also have different leverage. Washington can gate closed American models before release or during deployment, but it cannot retrieve weights already published. Beijing could preserve open access to capable mid-tier systems while placing its strongest models behind APIs or geographic controls, retaining some international influence without distributing every frontier checkpoint.
AI model governance and compliance tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Open Releases Meet Policy Limits
Open-weight releases have arrived rapidly, with the source material counting four frontier-class releases in eight weeks. That cadence supported an ecosystem in which organizations could download models, operate them on local infrastructure and avoid dependence on a single commercial API.
Signs of a tiered Chinese approach were already visible before the July report. Qwen 3.6 was released openly, while Qwen 3.7 Max remained available only through an API, according to the source material. A May roundtable published in a Supreme People’s Court journal also examined tiered governance for open-source AI. Separately, the Manus acquisition was unwound in June amid broader cross-border investment rules.
“The licensing generosity is a policy, not a law of nature.”
— Thorsten Meyer AI
AI-Powered Cybersecurity: AI Tools for Enterprise Security | AI for Network Security | AI Risk Management | AI in Cyber Policies | Cyber Threat Management AI | ML in Fraud Prevention
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Future Chinese Releases Remain Unsettled
It is not yet clear whether China will adopt formal restrictions, which models would be covered or whether rules would apply to published weights, unreleased systems or API access. The July 7 account concerned policy discussions rather than a final order, and no enforcement timetable was provided.
Questions also remain about the reach of the US regime. The reported severity of the jailbreak finding behind the Anthropic controls was not independently confirmed, and the restrictions were later removed. It is unknown how often officials will use similar authority or how the August 1 provisions will affect launches beyond GPT-5.6 Sol.
AI model monitoring and review platforms
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Launch Terms Become the Key Test
The clearest signal will come from the release conditions attached to the next generation of models. Observers will be watching whether systems such as a future GLM-6 appear with downloadable weights at launch, arrive only through controlled APIs or carry geographic limits.
Organizations dependent on foreign models can reduce exposure by archiving lawfully available MIT- or Apache-licensed weights, tokenizers and inference software; testing current models against real workloads; and building systems that can route requests across local and hosted options. European teams will also need to price policy risk alongside compute and operating costs as the United States and China define their model-access rules.
Key Questions
Are existing open-weight models being removed?
No broad removal is confirmed. Weights already published and lawfully downloaded generally remain available through existing copies and mirrors. The larger concern is whether laboratories will release future frontier generations under similarly permissive terms.
Has China formally restricted overseas model access?
Not based on the supplied reporting. Reuters described MOFCOM discussions with technology companies, not a final decree. The scope, timing and legal form of any restrictions remain unknown.
What happened to the US controls on Anthropic models?
The Commerce Department imposed controls on Claude Fable 5 and Mythos 5 in June, according to the source material, then lifted them after a weeks-long dispute. The episode showed that access to a deployed closed model can be interrupted by government action.
Why is this especially relevant to Europe?
Many European local-first AI projects rely on American or Chinese models rather than domestic frontier systems. Changes to release policy could raise costs, slow upgrades and leave public or commercial services exposed to decisions made abroad.
What indicator should organizations watch now?
The main indicator is how new models are released: downloadable weights, controlled API access, customer approval or geographic restrictions. Release conditions may reveal more about long-term availability than benchmark scores alone.
Source: Thorsten Meyer AI
