US Export Order Forces Anthropic to Halt AI Models Globally in 90 Minutes

In an unprecedented move, the US government issued an export-control directive to AI startup Anthropic, requiring it to disable access to its flagship Fable 5 and Mythos 5 models worldwide. The company stated it had approximately 90 minutes to comply, leading to a sudden shutdown that disrupted users and businesses across the globe.

The directive, received on June 12, cited national security concerns. Authorities believed there might be ways to bypass safeguards and use the models to identify software vulnerabilities. Anthropic argued that the concerns stemmed from a narrow and disputed 'jailbreak' scenario, but chose to shut down access for all users to ensure compliance.

This incident marks one of the first major instances where export-control measures have directly disrupted access to frontier AI models, rather than the hardware that powers them. Developers, enterprises, researchers, and paying customers worldwide lost access almost instantly. Even some of Anthropic's own employees were affected.

Industry observers say the event serves as a stress test for the rapidly growing AI economy. Many companies have embedded AI models into critical operations—customer support, software development, analytics, content production, and decision-making. The shutdown disrupted entire business processes that were redesigned around AI-assisted execution.

Chandrakant Agrawal, Co-Founder and CEO of AppSquadz, noted that the incident should be viewed as a business continuity issue. 'This incident highlights that AI has evolved beyond a technology decision and is now closely tied to regulatory, geopolitical, and business continuity considerations,' he said. He added that organisations depending heavily on a single AI provider expose themselves to operational risks that can surface without warning.

The implications are particularly significant for India's IT services industry, which has invested heavily in AI-led transformation projects for global clients. Many solutions rely on advanced models from a handful of US companies. Agrawal said clients increasingly expect consistent outcomes regardless of changes in the underlying technology ecosystem. Technology firms must now build resilient AI platforms that can adapt to changing providers, regulations, or market conditions.

The event underscores the need for businesses to diversify AI dependencies and prepare for sudden regulatory changes. As AI becomes more embedded in critical operations, the lesson is similar to how companies approach cloud infrastructure: diversify to mitigate risk.