On July 1, 2026, Anthropic PBC announced that it would restore access to two of its most powerful LLMs, Mythos and Fable. The models had been pulled worldwide after a US export order was issued on June 12, 2026.
Key Developments
- Access to Mythos and Fable resumes in the evening of July 1, 2026.
- The ban followed reports that Indian organisations had obtained Mythos for cybersecurity research.
- Anthropic added a safety classifier that redirects blocked requests from Fable 5 to Opus 4.8, its earlier‑generation model.
- India’s government is considering equity support for Sarvam AI to reduce reliance on foreign LLMs.
- A senior official reiterated at the Pax Silica summit that once access is granted, it should not be withdrawn.
Important Facts
• Mythos is described as capable of scanning large codebases to uncover vulnerabilities that have eluded human reviewers for decades.
• Fable also performs code analysis but blocks biology‑related queries to prevent the creation of bioweapons.
• The safety system notifies users when a request is blocked and automatically forwards it to Opus 4.8, the model that was widely available in May‑June 2026.
• The Indian government views AI as a strategic tool for “maximum utility” and broad impact, contrasting with Western concerns about job displacement.
Exam Relevance
The episode illustrates the intersection of technology governance and international security. Candidates should note:
- How export controls can affect domestic AI development and strategic autonomy (GS2: Polity).
- The role of sovereign AI startups like Sarvam AI in reducing dependence on foreign LLMs.
- Implications for cybersecurity policy, given cybersecurity capabilities of models like Mythos.
Way Forward
India should:
- Formulate clear AI policy frameworks that balance innovation with security.
- Encourage domestic R&D through equity stakes in firms such as Sarvam AI.
- Maintain dialogue with the United States to align export regulations with India’s strategic needs.
- Invest in safety classifiers and monitoring mechanisms to prevent misuse of powerful LLMs.
By addressing these areas, India can harness AI’s benefits while safeguarding national security and technological sovereignty.