Source-led article
Anthropic’s Fable 5 and Mythos 5 Suspension Sparks Sovereign AI Debate in India
Anthropic has suspended access to its newly launched Claude Fable 5 and Mythos 5 AI models globally, including for non-US nationals, just three days after their release. This action follows an export control directive from US authorities, which has reignited the debate in India about its dependence on foreign AI technologies and the necessity for a stronger national AI strategy.
The suspension highlights how advanced AI systems are increasingly viewed as technologies with national security implications, prompting calls from Indian industry leaders for accelerated domestic AI development.
US Directive and Anthropic’s Response
Anthropic announced the suspension after receiving an order from US authorities to block access to Fable 5 and Mythos 5 for all foreign nationals. The company complied, effectively disabling the models for customers worldwide. Anthropic stated that the directive offered few details regarding the national security concerns behind the move.
Based on its discussions with officials, Anthropic believes the issue stems from a technique that could “jailbreak” Fable 5, allowing users to bypass certain safeguards. However, Anthropic downplayed the significance of this finding, arguing that similar capabilities are available in other models, including OpenAI’s GPT-5.5, and are routinely used by security defenders. The company expressed disagreement that a “narrow potential jailbreak” should lead to recalling a commercial model deployed to millions, suggesting such a standard would halt new model deployments across the industry. This is not the first time Anthropic has faced US scrutiny; earlier this year, the Pentagon reportedly designated it an “unacceptable supply chain risk.”
The Launch of Fable 5 and Mythos 5
Anthropic had introduced Fable 5 and Mythos 5 on June 9, describing them as significant advancements in AI capabilities. Fable 5 was designed as a general-purpose model within Anthropic’s new Mythos-class, which the company claims surpasses its Opus family in areas like software engineering, scientific research, visual reasoning, and complex problem-solving.
Mythos 5, a version with some cybersecurity safeguards lifted, was initially made available only to organizations with access to Claude Mythos Preview under Project Glasswing. This project, focused on cybersecurity, operates in collaboration with the US government. Earlier in the month, Project Glasswing expanded to include approximately 150 organizations across more than 15 countries, including India, to help identify and remediate software vulnerabilities.
Key Facts
| Feature | Detail |
|---|---|
| Models Suspended | Claude Fable 5, Claude Mythos 5 |
| Reason for Suspension | US government export control directive |
| Anthropic’s Stance | Disagrees with significance of “potential jailbreak” |
| Impact on India | Reignites debate on dependence on foreign AI and national AI strategy |
India’s Sovereign AI Debate Intensifies
The suspension has amplified a long-standing discussion in India about reducing its reliance on foreign AI platforms and developing a more effective national AI strategy. India’s government is already pursuing the ₹10,372 Cr IndiaAI Mission, which includes subsidised AI compute and startup financing. However, investors and founders are urging for more decisive action.
Sridhar Vembu, founder of Zoho, emphasized the link between technology and national power, stating, “National sovereignty, national security, all of it is now about technology.” He added that “globalisation is dead and Bharat must find her own way ahead.” Hemant Mohapatra of Lightspeed Venture Partners described the event as a “sovereign AI is real” moment. Mohandas Pai, chairman of Aarin Capital, called for a significant national AI push, proposing an annual fund of ₹50,000 Cr for deeptech and AI investments.
Rethinking AI Product Development
The incident is also prompting Indian startup founders to re-evaluate their approaches to building AI products. Saurabh Awasthi, cofounder of Augmen.IO, cautioned against over-dependence on single frontier AI providers. He described the suspension as an “alarm” for both startups and nations, advocating for resilience through multi-model architectures, open-source fallback models, and on-premise deployments.
Conversely, Shayak Mazumder, founder and CEO of Adya.ai, viewed the suspension as a “knee-jerk reaction” by the US government and likely temporary. He argued that while Fable 5 is strong for long-horizon tasks, its advantages over models like GPT-5.5, Gemini, and Anthropic’s Opus series are often overstated. Mazumder suggested that India should focus on building smaller models, creating fine-tuning studios, and developing well-harnessed models rather than striving for its own large foundational models.