On June 12, at 5:21 PM Eastern Time, Anthropic received a directive it did not expect. The US Department of Commerce ordered the company to immediately suspend access to its two newest and most capable AI models — Fable 5 and Mythos 5 — for every foreign national on earth, including Anthropic’s own non-American employees. The company had no practical choice. It killed the models for everyone.
Fable 5 had been live for exactly three days.
In the hours that followed, a policy debate that had long lived in Indian think-tank reports and government committee notes became an operational reality. Indian builders who had already wired Fable 5 into production workflows woke up on Saturday morning to a dead API. The abstract question of AI sovereignty had acquired a specific, dated, documented shape: June 12, 2026, 5:21 PM ET.
That was two weeks ago. The ban is still in force. And the story has changed significantly since.
What Anthropic Was Told — and What the NSA Actually Did
The original framing from the US government centred on a jailbreak — a technique for bypassing Fable 5’s safety guardrails. Anthropic disagreed publicly, arguing it had received only “verbal evidence of a potential narrow, non-universal jailbreak” and that the capability involved was already present in other publicly available models.
“The level of capability displayed there is widely available from other models (including OpenAI’s GPT-5.5), and is used every day by the defenders who keep systems safe.”
— Anthropic, official statement, June 12, 2026
But testimony that emerged from a Senate Intelligence Committee briefing told a different story. NSA Director Gen. Joshua Rudd told Senator Mark Warner that Mythos 5 had autonomously breached nearly all NSA classified systems in a red-team exercise — not over weeks, but in hours. This was not a theoretical jailbreak risk. It was a live demonstration of what the model could do when pointed at classified infrastructure by trained operators.
A second layer of context also emerged. A White House Executive Order issued on June 2 — ten days before the suspension — mandated a 30-day government pre-release review framework for all frontier AI models. Anthropic launched Fable 5 on June 9, seven days after that order, with no pre-brief to the government. The directive may have been as much about enforcing that framework as about the alleged jailbreak. The two are not mutually exclusive.
Anthropic warned that if this standard were applied consistently, it “would essentially halt all new model deployments for all frontier model providers.” It called the directive a misunderstanding and said it was working to restore access. Two weeks on, access has not been restored.
What Changed in Two Weeks
Several significant developments have reshaped the picture since the suspension landed.
Fable 5 appears to be coming back — partially
On June 21, Fable 5 reappeared in the Claude Android app’s model selector. Users attempting to access it received error messages reading “server is temporarily rate-limiting requests” rather than the previous “model unavailable” — a meaningful distinction suggesting a live back-end rather than a hard block. Prediction markets placed roughly 57% odds on full restoration before July 1.
Anthropic has an identity solution in the works
The company’s new identity verification policy, taking effect July 8, will collect government ID, a facial photo, and biometric templates from users via identity platform Persona. This would allow Anthropic to restore Fable 5 for verified US persons without waiting for the Commerce Department directive to be formally lifted — sidestepping the regulatory blockage through a compliance architecture rather than a policy reversal.
Trump signalled a thaw at G7
President Trump met Anthropic CEO Dario Amodei at the G7 summit in Évian-les-Bains and indicated he had eased national security concerns about the models. The Commerce Department directive remains legally in force, but political intent appears to have shifted. The Information reported that the White House is unlikely to extend similar restrictions to other AI companies and is privately attributing the Anthropic situation to the company’s own handling of the alleged jailbreak vulnerabilities.
The open-source market moved immediately
When Fable 5 went dark, enterprises needed alternatives within hours. Four open-weight coding models stepped into the gap before Anthropic could restore access: Cohere’s North Mini Code, Moonshot’s Kimi K2.7-Code, Zhipu’s GLM 5.2, and a fourth from Mistral (The New Stack). The Fable 5 ban effectively became a live advertisement for open-source AI resilience — exactly the argument India’s sovereignty advocates have been making for years.
India Was Deeply Embedded, Not Casually Exposed
When the suspension landed, coverage initially treated India as one of many affected markets. The data tells a different story. Anthropic and OpenAI have both described India as their second-largest market after the US — reflecting years of enterprise investment, developer adoption, and institutional partnerships.
TCS had been training 50,000 employees on Anthropic’s models at the time of the suspension. Infosys had active collaborations underway. Select Indian institutions had only weeks earlier gained access to Mythos 5 through Project Glasswing — Anthropic’s government-linked cybersecurity programme — before that access was revoked along with everyone else’s. Anthropic announced the TCS partnership in the same statement as the suspension, in what became one of the more ironic juxtapositions in recent Indian tech history.
Saket Dandotia, co-founder and CEO of Onetab.ai, told CNBC that the suspension had come close to breaking his business of building AI applications for enterprises — saved only by multi-model diversification he had built in advance. But he was clear about the limits of that strategy.
“Diversification buys time; it doesn’t buy independence.” — Saket Dandotia, Co-founder & CEO, Onetab.ai
An ADP Research report released in the same week found that 41% of Indian workers use AI nearly every day — higher than 26% in China and 19% in the US. India’s workforce dependence on AI tools is now greater than that of any other major economy. The infrastructure underpinning that dependence is almost entirely foreign.
Sridhar Vembu Said What Many Were Thinking
Zoho founder Sridhar Vembu was among the first Indian voices to respond publicly, posting on X within hours of the suspension.
“Technology is the ultimate weapon. National sovereignty, national security, all of it is now about technology. Globalisation is dead and Bharat must find her own way ahead.” — Sridhar Vembu, X, June 13, 2026
Vembu argued for a pivot toward smaller models — both Indian-built and Chinese open-source — and questioned the logic of paying providers who can, at a foreign government’s direction, simply stop selling to you. He urged Indian organisations to embrace open-source alternatives immediately, rather than waiting for a policy solution that may not arrive on a useful timeline.
Mohandas Pai Goes Directly to the PMO
Mohandas Pai, chairman of Aarin Capital and former CFO of Infosys, went significantly further than Vembu — and significantly further than his initial reaction, which this publication reported when the suspension first landed.
Pai posted on his official X account, tagging Prime Minister Narendra Modi, along with Union Ministers Ashwini Vaishnaw, Nirmala Sitharaman, Piyush Goyal, and Amit Shah, BJP IT head Amit Malviya, and the Reserve Bank of India — a unusually public and specific demand for government action from one of India’s most senior tech-industry voices
“Existing govt programs are too slow, way too small to make any large impact. We need an annual 50000 cr fund for deep tech and AI, a 200,000 cr ELGS Guarantee Fund to build Hyper cloud, hardware and chips.” — Mohandas pai, X, June 13, 2026
He proposed a formal “India AI Mission” led by both public and private sector representatives, with Nandan Nilekani — Infosys co-founder and architect of Aadhaar — as vice-chair. The financial ask was not small: an annual ₹5 lakh crore deep-tech and AI fund, plus a ₹2 lakh crore guarantee fund for cloud infrastructure, hardware development, and semiconductor manufacturing.
His assessment of India’s current position was blunt. Existing government programmes were described as “too slow, way too small to make any large impact.” He argued that India risks falling behind in the global AI race in a way that will be structurally difficult to reverse — and that the Fable 5 suspension had made the timeline for action shorter, not longer.
Hemant Mohapatra said “sovereign AI is real”
Hemant Mohapatra, Partner at Lightspeed Venture Partners India, summarised the investor community’s reaction in fewer words on X.
The “sovereign AI is real” moment is here. Nation-states will soon start needing citizenship and/or security clearances to work on the next SOTA models the way they do for defense, space, nuclear tech. It is only a matter of time. Talent wars here will be crazy. — Hemant Mohapatra, X, June 13, 2026
The Structural Gap India Has Not Closed
India’s sovereign AI position in mid-2026 is best described as ambitious infrastructure, real compute investment, and a frontier model gap that is not closing quickly enough. The IndiaAI Mission’s ₹10,372 crore budget and 38,000+ onboarded GPUs represent genuine public investment in compute access (PIB). Sarvam AI‘s flagship model has over 100 billion parameters. Krutrim has committed $230 million and announced India’s largest planned supercomputer in partnership with Nvidia.
But Sarvam’s model is still orders of magnitude smaller than Fable 5’s architecture. And current sovereign AI models in India are built on Nvidia hardware — which itself is subject to US export controls. The compute sovereignty problem runs deeper than the model sovereignty problem, and both remain unsolved.
“Even if this is corrected or reversed, the Anthropic episode shows there’s no such thing as a geopolitically neutral foreign LLM. American AI models are bound to American geopolitics.” — Prasanto Roy, a New Delhi-based technology policy expert said to TechCrunch
HCL Tech‘s ₹14.27 billion investment in Sarvam — which CNBC noted was less than 10% of what the company paid shareholders in dividends in FY26 — captures the scale mismatch. India’s private sector is investing in domestic AI, but at a fraction of what the problem requires.
The Honest Near-Term Answer Is Open Source
While the long-term structural debate continues, the practical lesson of the Fable 5 suspension is already being absorbed. Enterprises that survived the outage without operational damage were those running multi-model architectures with open-source fallbacks. Those that had concentrated their workflows on a single frontier provider lost access to those workflows the moment Washington sent a letter.
The four open-weight models from non-US artificial intelligence companies stepped into the gap — Cohere, Moonshot, Zhipu, and Mistral — demonstrated that open-source alternatives exist, are capable, and are not subject to US export control directives. For Indian enterprises building on AI infrastructure, this is not an abstract point. It is a procurement decision.
Vembu’s prescription — smaller models, open-source fallbacks, reduced single-provider dependence — is the correct near-term response. Pai’s ₹5 lakh crore annual fund is the correct long-term response. The White House’s private signals suggest the Fable 5 directive may be unwound before July. But the fact that it happened — that access to the world’s most capable AI model was terminated for 1.4 billion people in under 90 minutes, by a single letter, from a government that consulted none of the affected countries — is not a misunderstanding. It is the system working exactly as designed.
India now knows what that system looks like from the outside.
Frequently Asked Questions
What are Anthropic Fable 5 and Mythos 5?
Fable 5 and Mythos 5 are Anthropic’s most advanced AI models, launched on June 9, 2026. Fable 5 is the general-purpose public model with strong safety guardrails, capable of complex software engineering, scientific research, and long-horizon autonomous tasks. Mythos 5 shares the same architecture but with some safety constraints adjusted for cybersecurity use — it was restricted exclusively to approved organisations through Anthropic’s Project Glasswing programme.
Why did the US government suspend access to Fable 5 and Mythos 5?
The US Department of Commerce issued an export control directive on June 12, 2026, initially citing a potential jailbreak vulnerability in Fable 5. Subsequent Senate testimony from NSA Director Gen. Joshua Rudd revealed that Mythos 5 had autonomously breached nearly all NSA classified systems in a red-team exercise in hours — significantly expanding the stated justification. A White House Executive Order mandating 30-day pre-release government reviews for frontier AI models, issued ten days before the suspension, is also seen as a contributing factor.
When will Fable 5 access be restored for India?
As of June 23, 2026, access has not been formally restored. Fable 5 reappeared in the Claude Android app on June 21 with rate-limiting errors rather than hard blocks, suggesting a partial back-end restoration is underway. Anthropic’s identity verification policy taking effect July 8 — requiring government ID and biometrics via Persona — is expected to enable a compliant restoration for verified users. President Trump signalled a thaw at the G7 summit, but the Commerce Department directive remains legally in force.
What is India’s sovereign AI position after the Fable 5 suspension?
India has made real progress — the IndiaAI Mission has deployed 38,000+ GPUs, Sarvam AI is building large Indian-language models, and Krutrim has committed $230 million to frontier model development. But no Indian model currently matches Fable 5’s capabilities in software engineering, scientific reasoning, or autonomous task execution. India’s sovereign AI models also depend on Nvidia hardware subject to US export controls, meaning the compute sovereignty problem runs alongside the model sovereignty problem.
What did Mohandas Pai propose to PM Modi after the suspension?
Aarin Capital chairman Mohandas Pai tagged Prime Minister Modi and six senior ministers on X, calling for a formal India AI Mission with Nandan Nilekani as vice-chair. He proposed an annual ₹5 lakh crore fund for deep tech and AI development, and a separate ₹2 lakh crore guarantee fund for cloud infrastructure, hardware, and semiconductor manufacturing — describing existing government AI programmes as “too slow, way too small to make any large impact.”
