The US has lifted export controls on Anthropic’s latest Fable 5 and Mythos 5 models, less than three weeks after ordering the company to suspend access over national security concerns. Anthropic said the Commerce Department removed the restrictions after new safeguards were put in place, allowing the company to start restoring access again.
The original June 12 order forced Anthropic to immediately restrict access for foreign nationals, but the company said it had no practical way to verify nationality in real time, so it ended up disabling both models for all users. That made the move much broader than a normal overseas export restriction and turned it into a full shutdown.
Why the US Stepped In
According to iTnews, Washington has stepped up oversight of new model releases amid concerns that advanced AI systems could be misused by military intelligence in countries such as China and Russia. The restrictions on Anthropic were part of that wider push.
Anthropic said the June 12 order followed Amazon researchers identifying a way to bypass Fable 5’s safeguards, allowing the model to identify software vulnerabilities and, in one case, generate code showing how a vulnerability could be exploited. That appears to have been one of the key triggers for the clampdown.
What Anthropic Changed
Anthropic says it has now implemented a new safeguard that blocks the behaviour described in that report. If a request gets blocked, it will instead be routed to the company’s Opus 4.8 model. Anthropic acknowledged that this may frustrate some users, but said it made the trade-off to keep Fable 5’s other capabilities more widely available.
The company also said it is deepening its work with the US government. That includes giving designated government partners expanded early access to both models and working with Amazon, Microsoft, Google and other Glasswing partners on shared standards for assessing and fixing jailbreaks, including a system for ranking their severity.
Mythos and Fable Are Back, But Not Without Strings
Mythos 5 had already been partially allowed back last week, but only for some “trusted” US organisations. iTnews says the model had earlier been made available to a broader group of companies through Anthropic’s Glasswing project, which was designed around finding cyber security vulnerabilities.
Now that the broader controls have been lifted, Anthropic says it is working with the US government to expand access to Mythos 5 to the wider domestic and international set of Glasswing partners. Fable 5, which Anthropic says is intended for the general public and has stronger safeguards, was set to become available again from Wednesday.
But the US has not walked away entirely. In a letter seen by Reuters and cited by iTnews, Commerce Secretary Howard Lutnick said the department reserves the right to reimpose a licence requirement if circumstances change or if Anthropic fails to meet its commitments.
This Is Bigger Than One Model
The broader significance here is that frontier AI releases are starting to look more like national security events than ordinary product launches. iTnews reports that the Trump administration has created a voluntary framework allowing developers to offer “covered frontier models” to the US government for up to 30 days before releasing them to trusted partners. It also notes that OpenAI recently delayed a full public launch of GPT-5.6 at the US government’s request, initially limiting access to a small set of vetted partners.
That suggests Anthropic’s clash with Washington was not a one-off. It looks more like part of a new pattern in which the most powerful AI systems are being scrutinised not just for commercial or consumer risks, but for geopolitical and cyber-security implications too. That final point is an inference from the reported framework and the Anthropic and OpenAI cases.
Why this matters for Australia
This matters for Australia because access to the most advanced AI models is increasingly shaped by US policy decisions, not just by what the companies themselves want to release. Australian businesses, researchers and developers using global AI tools can end up affected by export controls, licensing rules and security reviews even when those decisions are made offshore.
It also shows where the AI debate is heading. The big question is no longer just which company has the smartest model. It is also who gets access to those models, under what safeguards, and how quickly governments are willing to step in when they think the risks are too high. That trend is strongly suggested by the Anthropic restrictions and the new US framework described in the report.
The bigger takeaway is simple: Fable 5 and Mythos 5 are back, but the episode shows how quickly frontier AI access can now be tightened or reopened by government intervention.
Source links
Source: iTnews
