Rafomac News, Tech & Trends That MatterThe US government has suddenly tightened control over the world’s most powerful AI models, effectively deciding who can access cutting-edge technology. This move isn’t just about safety—it signals a shift in global AI power and sovereignty.
Anthropic’s Fable 5 and Mythos 5: Locked Down for Safety or Control?
It all began when Anthropic released two highly advanced AI models: Fable 5 and Mythos 5. At first, the launch went smoothly. But late on June 12, the Trump administration slapped export controls on these models, citing safety concerns. The result? Anthropic had to suspend access indefinitely.
Now, Fable 5 and Mythos 5 are back—but not as free as before. Their safeguards have been broadened so extensively that even innocent questions about topics like bioterrorism trigger blocks. One test query about AI research into bioterrorism was immediately flagged and shut down by Fable 5’s safety filters.
Meanwhile, Mythos 5, which had been available to select companies for cybersecurity projects under “Project Glasswing,” saw its partner list shrink sharply. Only a handful of US-based companies retained access, all handpicked in collaboration with the US government.
The US Government as Gatekeeper of AI Access
These developments suggest a new de facto licensing regime is emerging. The US government now effectively decides who can use the world’s most sophisticated AI technology. And Anthropic isn’t the only company facing these restrictions.
OpenAI recently announced new models too. CEO Sam Altman highlighted the upside—that the models are powerful and promising. But he immediately tempered the excitement by revealing that access would be launched in “limited preview” due to US government requests. Like Anthropic’s tech, this limited rollout means only select US companies can use the models.
Altman himself expressed unease about this arrangement, saying he supports safety checks but dislikes the government choosing customers. That tension reveals the friction between innovation ambitions and regulatory control.
AI Companies Propose Radical Solutions to Circumvent Restrictions
AI companies are pushing back hard. To regain freedom to deploy their technology globally, they’ve floated bold ideas—like creating an AI sovereign wealth fund for the US. Reports suggest that some companies are even considering giving the government a 5% stake, potentially valuing billions of dollars, to ease access restrictions.
This is more than a business issue. Nations are scrambling to align their economies and militaries with AI capabilities, yet don’t control who accesses these breakthroughs. The power stakes are high.
The G7 and Global AI Power Dynamics
A recent G7 meeting put AI in the spotlight. Three major AI leaders—Sam Altman, Demis Hassabis, and Dario Amodei—presented ideas on managing AI model releases. While world leaders listened, the real decision-making epicentre remains the US administration.
Geopolitics play a huge role: the US and China lead in AI intelligence, with European players like the French company Mistral notably marginalized. The UK and EU’s absence from decisive conversations signals a growing gap in global AI influence.
What Does AI Sovereignty Really Mean Now?
The term “sovereignty” might sound abstract, but here it boils down to control—who gets to use and profit from AI technology. The UK government is trying to boost its agency in AI, but real power flows where the leading AI companies and governments reside, especially in the US and China.
This new era of AI control raises critical questions about innovation, security, and the future balance of global power. It’s no longer just about building smarter machines—it’s about who holds the keys.
For anyone curious about AI’s evolving landscape, understanding these dynamics is crucial. And seeing these shifts happen in real time makes it clear that the AI race is as much political as it is technological.
