Anthropic pulling the plug will push Indian IT firms to adopt model-agnostic architectures, AI fallback plans

With US restrictions exposing the risks of relying on a single AI models for client use cases, Indian IT firms will need to adopt open-source alternatives and build flexible systems to de-risk and seamlessly switch between different AI models.

On June 12, Anthropic received a directive requiring the company to suspend access to Fable 5 and Mythos 5 for all foreign nationals, whether they are inside or outside the United States, including foreign-national employees working at the company.

This came barely a day after India’s biggest IT services giant Tata Consultancy Services (TCS) announced its strategic partnership with Anthropic to roll out Claude assistants for 50,000 employees. Infosys too has a partnership with the company to co-develop industry use cases, which was announced in February.

Indian IT firms’ business models involve building broader ecosystem partnerships across companies such as OpenAI, Anthropic, Google, Meta, Microsoft, AWS among other AI players — making it an equal playing field for all with no decisive advantage to a particular AI model.

According to Satish Viswanathan, former MD at Accenture India, in the recent AI-focused client engagements, the negotiations started with the business problem, not the model. The models get selected later based on the use case, cloud architecture, data environment, security posture, and client preference.

But, with the new restriction from Anthropic, this logic changes.

“Model substitution will become an engineering and governance challenge, not just a procurement decision. Firms may now need to test the same use case across multiple models, build fallback options, redesign prompts and RAG pipelines, and create abstraction layers for agent workflows and safety controls,” said Viswanathan.

The lesson for services providers is that they can no longer build their AI strategy around a single model ecosystem.

Phil Fersht, founder and CEO of HFS Research believes the “winners” will be firms that develop model-agnostic architectures and maintain expertise across multiple platforms, including Anthropic, OpenAI, Google, Meta, Mistral, and emerging sovereign alternatives.

“The value is increasingly shifting away from the model itself and toward orchestration, governance, workflow integration, data engineering, and domain expertise,” he told Moneycontrol.

Consequently, there will be a growing demand for open-source models, sovereign AI initiatives, regional AI ecosystems, and architectures that allow organizations to switch models without redesigning applications.

Meanwhile, Indian IT firms have already started to take sides and double down on enabling sovereign frontier models. Case in point, HCLTech is set to invest $150 million into India’s Sarvam AI at a $1.5 billion valuation, Moneycontrol exclusively reported earlier citing sources.

Project delays, risk reassessments

While IT companies are unlikely to abandon leading proprietary models, this crisis will open up opportunities for the industry to move from “model implementation partners” to “AI resilience architects.”

“The near-term impact will be project delays, contract reviews, and risk reassessments for solutions relying on the most advanced Claude tiers,” Iti Choudary, Senior Analyst, Everest Group told Moneycontrol.

She added that the larger shift will be architectural: service providers will need to design AI solutions with model portability, abstraction layers, fallback models, and jurisdiction-aware deployment as standard features.

Perhaps the silver lining lies in the fact that for the vast majority of the work happening with AI models today, having the latest frontier models is not absolutely critical, noted Prashant Yadav, Partner for digital and technology at leadership consultancy firm Amrop India.

Yadav, however, added that if US continues such restrictions for a long time, the markets could end up “channeling more funding toward European and Chinese AI models, which may be counterproductive for the USA.”

At a country level, for India, this episode reinforces the strategic importance of building sovereign AI capabilities.

“While access restrictions are not a major operational issue today, they highlight the long-term risk of relying exclusively on foreign frontier models whose availability could be limited by geopolitical or regulatory decisions,” said Gaurav Parab, principal research analyst, NelsonHall.