Artificial intelligence company Anthropic has reestablished access to its advanced models, Fable 5 and Mythos 5. This decision comes after the United States government eased export controls, which had abruptly halted use of the systems for national security reasons, and the developer introduced new, more robust safeguards against potential intrusions.
Anthropic, one of the leading companies in the artificial intelligence sector, had previously suspended the availability of its innovative models, Fable 5 and Mythos 5. The interruption occurred earlier this month when the US government imposed export controls due to serious national security concerns.
The Fable 5 model will be offered to users across the planet from July 1st. The company confirmed the news in an official publication, indicating that access will be possible through its platforms, including Claude.ai and Claude Cowork.
In turn, access to Mythos 5 was restored to a restricted group of organizations in the United States. Release to these selected partners occurred on June 26, after receiving approval from the federal government.
Currently, Anthropic reported that it is coordinating with the US government to expand the availability of its models. The expansion seeks to reach a greater number of partners both nationally and internationally, within the scope of its initiative known as Project Glasswing.
“After several constructive conversations with the US government, we are relaunching the Fable 5 model with a new set of classifiers to identify and block more cybersecurity-related tasks,” the company stated.
Anthropic added in a post on Platform
India represents Anthropic’s second largest market, highlighting the country’s strategic importance for the company. Recently, the AI developer entered into an agreement with TCS to train 50,000 of its employees with the models, in addition to announcing a collaboration with Infosys to implement advanced enterprise AI solutions using the Claude model line. These partnerships highlight the potential for expansion and the need for clarity in export policies for Anthropic’s global operations.
Sources familiar with the matter revealed that some Indian government entities, as well as private companies, have recently gained access to the Mythos model. This access grant is part of a cybersecurity initiative known as Project Glasswing.
The US government’s initial decision to restrict access came to light after Amazon researchers identified a vulnerability known as a “jailbreak” in the system. This method allowed bypassing Fable 5’s security protections, enabling the team to detect flaws in the software and demonstrate how they could be exploited.
To resolve the issue, Anthropic collaborated closely with the US government. The joint work resulted in the training of an “enhanced security classifier” capable of detecting and blocking the specific bypass technique in more than 99% of cases.
Anthropic said it is intensifying its collaboration with the US government in areas of model testing and safeguards implementation. This joint effort provides early access to models and protection mechanisms for evaluation, sharing data on intrusions and misuse, and dedicating resources to cooperative research.
