Washington's Anthropic ban revives the case for sovereign AI in India
A US order pulling Anthropic's Fable 5 and Mythos 5 offline has 76 security veterans in revolt and Indian buyers asking who really controls the models they rely on.
The News
Anthropic has suspended worldwide access to its two newest and most powerful AI models, Fable 5 and Mythos 5, after the White House issued an export control order on 12 June 2026 demanding the company block all foreign nationals, including its own overseas staff, from using them.
The models had been live for barely three days. Anthropic launched them on 9 June, describing Fable 5 as a system whose capabilities exceed anything it had previously made generally available, with Claude Mythos 5 built on the same underlying model but with some safeguards lifted.
The clampdown has triggered a sharp backlash. On 15 June, a group of 76 cybersecurity practitioners published an open letter urging the administration to reverse course. Signatories include Alex Stamos, the former Facebook chief security officer, Bugcrowd founder Casey Ellis, cryptographer Jon Callas, internet pioneer Paul Vixie, Luta Security founder Katie Moussouris and SocialProof Security chief Rachel Tobac. The order reportedly followed a report from Amazon researchers, which the letter's authors argue never demonstrated a genuine jailbreak.
Why It Matters
The protest reframes a national-security move as an own goal for defenders. The signatories contend the restriction strips security teams of their sharpest tool just as attackers gain access to comparable systems elsewhere. Moussouris said asking a model to repair flawed code is "the most valuable thing an AI model can do for defensive security", and that locking it away weakens, rather than protects, software supply chains.
The deeper signal is about control. Washington has shown it can switch off a private company's flagship product overnight and dictate who may touch it. The last time a frontier lab pulled a marquee model under government pressure, the conversation stayed academic. This time it is operational, and every enterprise that wired its workflows into Claude over the past year now has a continuity problem it did not price in.
Indian Angle
For India, the order lands squarely on a nerve. Indian developers, IT services majors and startups have leaned heavily on US frontier models, and an export rule that bars foreign nationals by definition reaches into Bengaluru, Hyderabad and Pune. A bank or insurer that built customer-facing tooling on Claude could wake to find the dependency frozen by a decision taken in Washington with no Indian input.
That is the strongest argument yet for the IndiaAI Mission's push toward indigenous foundation models. Sarvam AI, Krutrim and a handful of academic efforts have promised sovereign capability for years; an episode like this turns that pitch from nice-to-have into board-level risk management. Procurement teams at Indian enterprises are likely to start writing model-portability and exit clauses into contracts, and to hedge across providers rather than standardise on one American lab.
Regulators will take note too. MeitY's data-localisation instincts and the RBI's caution on critical infrastructure dependencies now have a live case study. Expect renewed momentum behind domestic compute, open-weight alternatives and clearer rules on what counts as a resilient AI stack for regulated Indian sectors.
FAQ
When did the order take effect?
The White House issued the export control order on 12 June 2026, and Anthropic suspended access to Fable 5 and Mythos 5 over the following weekend. The models had launched only on 9 June, giving users barely three days before the shutdown.
Why are security experts opposed?
The 76 signatories argue the restriction hands an advantage to attackers while disarming defenders, who use such models to find and fix vulnerabilities. They also say the triggering Amazon report did not prove a real jailbreak, making the response disproportionate to the demonstrated risk.
What does this mean for Indian companies?
Indian firms relying on Claude face a continuity risk, since the order targets foreign nationals broadly. The practical takeaway is to diversify across providers, add exit clauses and evaluate indigenous options such as Sarvam and Krutrim for critical workloads.
Where can I read the original report?
TechCrunch covered the protest letter and its signatories in detail. The link to the full original coverage appears in the attribution paragraph below.
This story was reported by TechCrunch. Read the full original coverage at TechCrunch.