US Curbs Advanced Anthropic AI Models to India

The US government has enacted emergency export controls restricting access to Anthropic’s most sophisticated artificial intelligence architectures, Fable 5 and Mythos 5. The directive confines these technologies strictly to the US, abruptly terminating operational access for international developers, enterprises, and research organizations throughout India.

A report has sparked debate across the technology world after the US government imposed emergency export controls on Anthropic’s most advanced AI models, Fable 5 and Mythos 5. Under the directive, access to the models would be restricted to the US, making them unavailable to foreign nationals and organisations outside the country.

The restrictions were imposed due to national security and cybersecurity concerns related to the advanced capabilities of the Mythos system. Following the directive, Anthropic was forced to disable access to the models for foreign users, businesses and developers worldwide, including those in India.

The development has raised several questions about the capabilities of these AI systems and what their restriction could mean for countries such as India. Meanwhile, a bipartisan group of House lawmakers demanded answers from the Trump administration about why it imposed sweeping restrictions on Anthropic’s latest artificial intelligence models — and whether rival tech companies should expect similar treatment.

What are Fable 5 and Mythos 5?

Anthropic recently launched its next-generation “Mythos-class” AI architecture. The company offers the technology in two versions. Claude Fable 5 is the commercial version available to developers and businesses and includes extensive safety guardrails designed to block harmful requests.

Claude Mythos 5, meanwhile, is described as a more unrestricted version with many of the safety filters removed. It is reportedly available only to a small group of vetted researchers and cybersecurity professionals through a programme called Project Glasswing. Partners listed in the project include major US entities like Amazon Web Services (AWS), Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks.

According to the company, these models represent a major leap in AI capability, with the ability to work autonomously on complex engineering and coding problems for hours without losing focus. Fable can uncover zero-day vulnerabilities—undiscovered security loopholes that can be used to launch attacks before they are patched, similar to the historic WhatsApp vulnerability exploited by Pegasus.

Unlike older AI models that are designed mainly for short conversations and quick responses, the Fable and Mythos series is built for what Anthropic calls Endurance. In simple terms, the models can handle large and complicated tasks for extended periods without drifting off track.

According to the report, payment company Stripe used the technology to completely rewrite a 50-million-line software project in a single day, a task that would normally take a team of engineers several months to complete.

Another major feature is memory. The models reportedly have a 1-million-token context window, allowing them to read, remember and analyse thousands of pages of documents, financial statements, charts and reports at the same time.

Why did the US impose restrictions?

According to the report, researchers discovered a “jailbreak” technique shortly after launch. The method allegedly allowed users to bypass some of Fable 5’s safety protections through carefully crafted prompts.

The report claims the technique enabled the AI model to analyse software code and identify hidden vulnerabilities. Amazon’s CEO reportedly brought these findings to the attention of senior officials in the Trump administration.

Concerned that foreign governments or state-sponsored hackers could exploit the technology to develop automated cyberattack tools against critical infrastructure, the White House reportedly issued an emergency export-control order citing national security concerns. This move affirms strict sovereign control over frontier AI architectures.

Impact on India

The restriction on Fable 5 and Mythos 5 affects foreign nationals globally, including users in India.

For India, the impact could be significant, as several IT companies, technology startups and cybersecurity researchers had reportedly integrated Fable 5 into their workflows. The sudden cutoff blocked independent developers from utilizing the system to secure their vibe-coded applications.

In addition, because the restrictions reportedly apply to non-US citizens worldwide, some American technology companies operating Global Capability Centres (GCCs) in India may be required to block access to the models for their Indian employees.

The development has also reignited discussions around India’s technological self-reliance. Several industry leaders have argued that dependence on foreign AI infrastructure leaves Indian companies vulnerable to geopolitical decisions taken abroad. This mirrors past actions during Trump 1.0, when US curbs on Android OS security updates forced China to develop HarmonyOS.

Furthermore, critics argue Anthropic’s global commitments ring hollow. While entering partnerships with Tata Consultancy Services (TCS), Anthropic simultaneously withheld critical system access from Indian digital infrastructure giants like TCS and Infosys, which manage sensitive data within domestic passport, banking, and tax frameworks.

In response, the Nishikant Dubey-chaired parliamentary standing committee on IT is being urged to summon Anthropic representatives to define future access terms. Concurrently, India’s cybersecurity agency, CERT-In, has held emergency meetings with infrastructure providers to counter potential adversarial deployment of Mythos systems.

Future Outlook

The sudden embargo highlights a growing paradigm shift where frontier AI models are treated as critical national security instruments rather than open commercial products. For India, this disruption serves as an urgent catalyst to achieve true digital sovereignty.

Industry experts suggest that India must pivot away from a reliance on foreign proprietary platforms and follow localized strategies, such as Kerala’s widespread adoption of open-source software for critical public infrastructure. To prevent external kill switches, future deployments must be physically hosted on domestic soil under the exclusive control of domestic firms.

Moving forward, the state-backed IndiaAI Mission will require substantial optimization, including localized hardware provisioning and diversified model development beyond singular domestic platforms like Sarvam. Academic ecosystems will also need structural restructuring, allowing universities to directly monetize research and empowering professors to launch state-supported startups. By leveraging its position as the largest available consumer market for technology, India may look to implement strategic market-access conditions to guarantee technological reciprocity and safeguard its critical networks.

FAQs

What are Claude Fable 5 and Mythos 5?

They are Anthropic’s next-generation AI models built on the Mythos-class architecture. Fable 5 is the commercially available version equipped with safety filters, while Mythos 5 is an unrestricted model reserved for specialized security researchers via Project Glasswing.

Why did the US government restrict access to Anthropic’s new models?

The emergency export controls were triggered after a jailbreak method allowed the AI to autonomously identify zero-day software vulnerabilities. The White House intervened due to national security concerns that foreign actors could exploit the system to build automated cyberattack tools.

How do these restrictions affect tech workers and companies in India?

Indian developers, startups, and IT firms have had their access to Fable 5 revoked. Additionally, Indian nationals working at US-based Global Capability Centres within India may be legally blocked from accessing or developing with these specific models.

What is the context window size of Fable 5 and Mythos 5?

The models feature a 1-million-token context window. This massive memory capacity allows the system to process, remember, and analyze thousands of pages of text, data charts, and institutional documents simultaneously.