Why frontier AI model restrictions challenge global operating models 

On June 12, Anthropic received a US government directive requiring it to suspend access to Fable 5 and Mythos 5 for foreign nationals, including foreign-national employees. Anthropic stated that it had to disable access for all customers to ensure compliance. 

The official rationale was national security. According to Anthropic, the government was concerned that users could bypass Fable 5’s safeguards to access certain cyber capabilities. The company has disputed the severity of that risk. 

Update: Anthropic has since redeployed Claude Fable 5 globally after the temporary suspension. The model is back across major platforms, including Claude Platform and Claude.ai. 

However, this incident sets an uneasy precedent for the global technology and services industry. Access to frontier Artificial Intelligence (AI) models is becoming a geopolitical and regulatory dependency, rather than a standard technology decision. 

Reach out to discuss this topic in depth. 

Why this matters for the global technology and services industry 

Global operating models have long relied on a simple assumption: work can move across locations because the supporting technology moves with it. That assumption has enabled Global Business Services (GBS) organizations, Global Capability Centers (GCCs), Service Providers, and distributed technology teams to allocate work based on talent availability, cost, scale, business proximity, risk, and time-zone coverage. 

The recent US government directive challenges that assumption. Frontier AI is no longer behaving like standard enterprise software. Instead, governments increasingly view it as a strategic asset whose availability can change with little notice.  

A global bank, for example, may distribute financial crime and compliance across GCCs in India, Poland, and the Philippines, expecting every team to use the same AI tools for investigations, documentation, and reporting. A Bangalore GCC or Polish delivery center could be locked out of a model the US parent could still, in principle, use. 

This creates AI access asymmetry. Organizations may still have the right talent, business case, vendor contract, and workflow in place, but still be unable to put AI capability where the work is done. Quality becomes uneven, handoffs increase, work is reallocated under pressure, and Service-level Agreements (SLAs) come under strain. The same operating model starts performing differently across locations. 

Key implications for the market 

The implications extend beyond AI providers. Enterprises, technology vendors, service providers, and governments all need to reassess how AI access affects their operating models. 

1. Enterprise and GBS organizations: AI access becomes a delivery model constraint

For enterprises and GBS organizations, the key question is whether AI-enabled workflows can operate consistently across every delivery location. As organizations embed AI into operations across financial crime, claims, cybersecurity, procurement, software engineering, customer support, and analytics, model access becomes a key operational dependency. 

A workflow designed around one frontier model may perform well in one geography but degrade in another if access rules change. Leaders, therefore, need clear visibility into which workflows depend on a single model, which geographies use that model, and whether tested alternatives exist. 

Before scaling AI across critical workflows, enterprises should ask: 

• Which workflows and delivery locations depend on a single frontier model? 
• What happens if a GCC, offshore team, or regional delivery hub loses access? 
• Can prompts, evaluations, controls, and documentation move across models? 
• Would work need to be reallocated if access became restricted by geography, nationality, sector, or use case? 

2. Technology and service partners: model access can handicap delivery capabilities

The same risk applies to technology vendors and service providers. Enterprise AI adoption increasingly depends on System Integrators (SIs), consulting firms, hyperscalers, Software-as-a-Service (SaaS) providers, and managed service partners. These firms are packaging frontier AI into propositions, delivery methods, industry solutions, and managed services. 

Recent market activity points to a broader pattern. OpenAI’s Frontier Alliances with Accenture, Capgemini, BCG, and McKinsey show how frontier labs are working with services firms on strategy, workflow redesign, integration, and deployment. Anthropic’s partnership with TCS reflects a similar approach for regulated industries. Hyperscaler-led partnerships are creating sector-specific AI hubs, such as TCS’s Google Cloud Gemini Experience Center for Business, Financial Services, and Insurance (BFSI) clients in Bengaluru. 

These partnerships are becoming part of providers’ Go-to-Market (GTM) strategies and delivery architectures. If a provider’s offerings, playbooks, talent pools, or managed service workflows depend heavily on one model family, access disruption can affect solution delivery, regional coverage, client commitments, and the ability to scale AI offerings. 

Enterprises would need to ask which models their partners rely on, which delivery locations use them, whether those models can be substituted, and what happens if access changes. Technology providers will need stronger answers on portability, fallback strategies, regional availability, and business continuity. 

If frontier AI partnerships are becoming central to GTM and delivery, how resilient is that strategy when access to a preferred model is no longer guaranteed? 

3. Regional and national AI autonomy becomes a strategic priority

If governments can restrict access to frontier AI models, AI capability becomes part of the geopolitical technology supply chain. 

Most countries are unlikely to build a domestic equivalent of OpenAI or Anthropic from scratch. The barriers are too high across compute, chips, data centers, research talent, capital, and enterprise distribution. The more likely path is pragmatic autonomy: local hosting for sensitive workloads, trusted regional model partnerships, sovereign cloud options, open-weight models, and domestic or allied AI capacity for regulated use cases. 

Europe is already moving in this direction through the AI Continent Action Plan. India, Japan, the Middle East, and other markets building AI-enabled operating models will also need to determine where critical AI capabilities are hosted, who controls access, and which workloads require domestic, regional, allied, or open-weight alternatives. 

The market is unlikely to move away from US-based frontier AI in the near term. Leading models will continue to matter for high-value use cases, but enterprises, providers, and governments are likely to put more emphasis on diversified model portfolios, regional AI partnerships, sovereign cloud options, and fallback models. 

The difficult question ahead 

Frontier AI will remain central to enterprise adoption. What has changed is the assumption that every team, partner, and delivery location will always have the same access level. 

For enterprises, GBS organizations, providers, technology partners, and policymakers, AI adoption has become an operational resilience challenge. If a critical model becomes unavailable to a geography, user group, or delivery partner, how will workflows continue? How will organizations maintain SLAs, customer commitments, and business continuity?  

Part 2 of this blog examines practical strategies for strengthening model portability, validating fallback options, improving partner due diligence, and building more resilient AI operating models.  

If you enjoyed this blog, check out, Purple people: The great unlock for GBS – Everest Group Research Portal, which delves deeper into another topic relating to GBS. 

To discuss how AI access risk could affect global operating models, reach out to Rohitashwa Aggarwal ([email protected]) or Dhairya Lohani ([email protected]).