The US government has ordered Anthropic to suspend access to its flagship AI models, Fable 5 and Mythos 5, under an export control directive issued on June 12. Anthropic said the order applies to “any foreign national, whether inside or outside the United States”, including its own foreign-national employees. The company has consequently disabled both models for all customers worldwide, while access to its other AI models remains unaffected. 

Anthropic disputes the rationale behind suspension: Anthropic said the government linked the directive to concerns about a potential method of bypassing, or “jailbreaking“, Fable 5’s safeguards. However, the company said it was not given “specific details” of the national security concern and had received only “verbal evidence of a potential narrow, non-universal jailbreak”.

According to Anthropic, the demonstrated technique uncovered only “a small number of previously known, minor vulnerabilities” that were “relatively simple”, adding that “other publicly-available models are able to discover them as well without requiring a bypass”.

The company defended Fable 5’s safety measures, stating that its safeguards are “substantially more effective than those of any previously deployed model” and that no tester had found “a universal jailbreak”. It also reiterated that “perfect jailbreak resistance is not currently possible for any model provider” and said it had adopted a “defense in depth strategy” combining safeguards, monitoring and 30-day data retention to detect and mitigate attacks.

Company says other frontier models have the same capabilities: Anthropic argued that the reported capability was “widely available from other models (including OpenAI’s GPT-5.5)”. While complying with the order, it said it disagreed that “the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people”, warning that such a standard could “essentially halt all new model deployments for all frontier model providers”.

How people are reacting online:

• Sovereign AI wake-up call: Sarvam AI founder Pratyush Kumar said users should not “confuse access with ownership, or adoption itself as advantage”, warning that organisations leveraging technologies with “external control loops” must accept that they are vulnerable. He argued that the incident strengthens the case for “Sovereign AI”, where India can “use and improve AI systems within their own perimeters”.
• Former MP Vijayasai Reddy similarly called the move a “wake-up call”, warning that India “can have its access restricted at any time” and urging the government to accelerate investments in domestic frontier AI and compute infrastructure.
• AI becoming a strategic asset: Public policy commentator Ajay Purushotham Shetty argued that the export controls underscore an emerging reality that “frontier AI is now a strategic asset”. He said “Software Swarajya is no longer optional” and warned that countries that do not control their own AI infrastructure risk losing strategic autonomy. According to Shetty, “those who control intelligence infrastructure will shape the future”, making domestic research, compute and model development a national priority rather than a commercial choice.
• How AI regulation could look like at scale: Aaron Levie, CEO of Box, framed the episode as an “early peek into what AI regulation would end up looking like at scale”. He argued that the government would effectively gain “sole discretion over when a model can be released to the public”. Levie warned that such a framework could result in a “backlog of AI releases” and cause progress to “dramatically slow down”. While stressing that “AI safety is incredibly important”, he argued that restricting models at the foundation layer rather than targeting harmful applications risks stifling innovation across the sector.
• Hurts everyone by making all systems vulnerable: Nikhil Pahwa, founder and editor of MediaNama, questioned the logic of restricting access to a model that many developers were using to identify and patch vulnerabilities. He noted that Fable “comes with security safeguards to prevent misuse” and said he had personally used it to “identify and plug security issues” in applications. Pahwa argued that “the US government refusing to let non-US nationals protect themselves is deeply problematic in an interconnected internet world“, particularly when vulnerabilities in foreign systems can ultimately affect American users as well. “This hurts everyone by making all systems vulnerable,” he wrote.
• Open-source and market alternatives: Entrepreneur Upamanyu Acharya argued that the suspension is unlikely to slow AI adoption because users will simply switch providers. He described the development as “bullish for open weights, not bearish”, noting that when “one frontier lab” is removed, demand “reroutes to the next lab”. Drawing on his own usage patterns, Acharya said he runs significantly more workloads on open-weight models such as DeepSeek because they deliver roughly “90% of the output at a 1000x less the cost”. 
• The real sovereignty problem: While many commentators called for sovereign AI models, Acharya pointed out that India faces a much deeper challenge. “That, not the Fable ban, is the real sovereignty problem,” he wrote, referring to the fact that AI spending largely flows overseas. He argued that model weights are relatively easy to replicate, claiming that “a sovereign model is the one layer of this entire stack we could have cloned in a weekend”. The real dependencies, he said, lie in foreign-controlled infrastructure: “power, copper, fabrication, ports, rare earths”, where India has “zero position”. “Every input I pay for sits outside the country,” he added.
• Geopolitics of frontier technology: Entrepreneur Sandeep Manudhane described the suspension as evidence that frontier technologies are increasingly being “geo-weaponised”. He characterised the move as the US government exercising a technological “Kill Switch” and warned that countries without domestic capabilities remain exposed to foreign political decisions. Manudhane argued that India is already a “near-total American digital colony” and suggested that similar restrictions could theoretically affect other critical technologies in the future. 

Also read:

• Anthropic opens Mythos-class AI to the public, but keeps high-risk capabilities behind guardrails
• SEBI forms task force, orders immediate cybersecurity overhaul amid Claude Mythos concerns
• Claude Mythos puts India on alert: CERT-In, telcos, banks assess unprecedented cyber risks