The relationship between AI companies and governments just entered a new phase.
After weeks of uncertainty and behind-the-scenes negotiations, Anthropic has secured US government approval to restore limited access to its powerful Mythos 5 AI model. The decision marks a significant victory for the company, but it also highlights how closely governments are now scrutinizing advanced AI systems.
The approval comes after the Trump administration raised concerns that Anthropic’s latest models could pose national security risks if their safety guardrails were bypassed.
What Happened?
Just two weeks ago, the US government abruptly blocked Anthropic from allowing foreign nationals to access two of its most advanced models, Mythos 5 and Fable 5.
The restrictions forced Anthropic to suspend global access to both systems while discussions with US officials intensified.
Now, according to a letter from US Commerce Secretary Howard Lutnick, Anthropic has made sufficient progress in addressing government concerns.
As a result:
- Mythos 5 can once again be deployed, but only to a limited group of trusted partners.
- Fable 5 remains under restrictions and is still unavailable for general use.
- Discussions between Anthropic and the US government are continuing.
This partial clearance effectively ends the immediate standoff, though broader questions remain unresolved.
Why Was the Government Concerned?
The central issue was security.
US officials were worried that sophisticated users could potentially “jailbreak” the models, meaning they might bypass built-in safety systems and access dangerous capabilities.
Anthropic itself has acknowledged that:
- No AI company has yet achieved perfect jailbreak resistance.
- Frontier AI models are becoming increasingly capable in areas such as cybersecurity and scientific research.
- Preventing misuse entirely remains an unsolved challenge.
The government was particularly concerned because Mythos 5 possesses advanced cybersecurity capabilities that could potentially be misused if adequate safeguards fail.
Exactly what technical changes Anthropic made to satisfy regulators has not been publicly disclosed.
What Is Mythos 5?
Anthropic originally introduced Mythos 5 as a highly capable cybersecurity-focused AI model.
Unlike consumer chatbots, Mythos 5 was never intended for broad public access.
Instead, it was distributed through an invite-only initiative called Project Glasswing, which provided access to selected organizations focused on cybersecurity and infrastructure protection.
Roughly 200 organizations reportedly had access, including major technology companies such as:
- Apple
- Microsoft
- Nvidia
- Cisco
The model was designed to help identify vulnerabilities, strengthen defenses and improve cyber resilience.
Because of these powerful capabilities, Anthropic intentionally limited access from the beginning.
What About Fable 5?
Fable 5 is essentially a more publicly accessible version of the same underlying technology.
However, Anthropic introduced additional safeguards before releasing it.
The company restricted responses related to:
- Certain cybersecurity activities.
- Sensitive biological topics.
- Other potentially dangerous areas.
When users ask questions in these categories, Anthropic routes responses through alternative, less capable systems.
Despite these protections, US officials remain cautious, and restrictions on Fable 5 continue.
That sends a strong signal that governments may increasingly intervene even when companies voluntarily implement safety measures.
A Bigger Industry Shift Is Underway
This episode may prove to be one of the most important moments in AI regulation so far.
For the first time, the US government directly intervened in the deployment of frontier AI models at a major scale.
The implications extend well beyond Anthropic.
In recent days, OpenAI also announced a more limited rollout of its latest model, GPT-5.6, to selected government-approved partners before wider release.
The message from Washington appears increasingly clear:
The most powerful AI systems may no longer be launched solely at the discretion of private companies.
Governments want a seat at the table.
Anthropic’s Broader Tensions With Washington
The dispute over Mythos 5 did not emerge in isolation.
Anthropic has reportedly been engaged in broader disagreements with parts of the US government, particularly regarding military applications of AI.
The company has historically positioned itself as a more safety-focused AI developer, often advocating stronger safeguards than many competitors.
That stance has occasionally created friction with government agencies seeking broader deployment options.
The latest negotiations appear to have eased immediate tensions, but the underlying policy debates remain unresolved.
What Happens Next?
Anthropic says it is now working to restore Mythos 5 access to approved partners as quickly as possible.
At the same time, the company continues discussions with regulators in hopes of:
- Expanding access to Mythos 5.
- Securing approval for wider use of Fable 5.
- Developing a long-term framework for handling future disputes over frontier AI systems.
The broader AI industry will be watching closely.
Because what happens here could shape how every advanced AI model is launched in the future.
The era of “build first, release later” may be coming to an end.
And the rules governing frontier AI are only just beginning to be written.