The US government ordered Anthropic to suspend global access to its Fable 5 and Mythos 5 AI models. This came under an export control directive, with Anthropic arguing the cited "jailbreak" was minor and non-universal. The company warns this sets a dangerous precedent, potentially halting all future frontier model deployments.
How We Got Here
The US issued an export control directive on June 12, forcing Anthropic to disable its flagship models for all foreign nationals. This effectively shut down Fable 5 and Mythos 5 worldwide, including for Anthropic's own non-US employees.
The Numbers
- Anthropic received only "verbal evidence" of the alleged jailbreak, lacking specific details.
- The company stated the technique uncovered "a small number of previously known, minor vulnerabilities" it called "relatively simple".
- Anthropic argued other frontier models, including OpenAI's GPT-5.5, provide similar capabilities.
- Sarvam AI founder Pratyush Kumar called the incident a "Sovereign AI" wake-up call for India.
- Former MP Vijayasai Reddy urged India to accelerate investments in domestic frontier AI and compute infrastructure.
What Happens Next
🇮🇳 Why This Matters for India
For Bangalore-based AI founders building atop US models, this incident highlights the immediate and critical need for India to develop truly independent, domestically controlled frontier AI compute and models.
The Take
Frontier AI is now a strategic, controlled asset, much like nuclear technology. Nations investing heavily in 'Software Swarajya' will win; India's push for self-reliance in this tech stack just got a significant boost.
Source:
MediaNama ↗