Secure enterprise AI search for the second largest railway operator in the world

At a glance

Company

SNCF Group, France's national rail operator and one of the world's largest railway groups, with 150,000 employees across 170 countries

The problem

SNCF needed to build an internal agentic platform capable of supporting more than 150,000 employees while ensuring strong usability, security, control, and auditability.

The platform required scalable load management, strict governance over accessible web domains through whitelisting and blacklisting, and private cloud-to-cloud connectivity without exposure to the public internet.

Implementation

Linkup's /search endpoint connected to SNCF GPT via a private cloud-to-cloud link, with domain whitelisting and blacklisting configured at the API level to scope every query to approved sources

Key results

• Web search deployed across 150,000 employees
• Full control over sources
• Infrastructure that handles their global query volume without performance degradation or data processing
  

Introduction

SNCF Group is France's national rail operator and one of the largest railway groups in the world, with over 150,000 employees across 170 countries. When SNCF rolled out its agentic platform to give employees access to AI-assisted work, they needed a web search layer that matched the security and operational standards of a state-owned infrastructure operator, not one designed for general consumer use.

The challenge

Standard LLM search queries the open web without restriction. Employees can surface results from low-authority, irrelevant, or potentially sensitive domains. Further, search traffic routed over the public internet does not meet the infrastructure security posture expected of a large state-owned operator like SNCF managing critical national rail infrastructure.

Pain points before Linkup:

• No domain control: Search operated across the entire public web, with no way to enforce which sources were allowed or blocked. SNCF had no mechanism to guarantee the reliability, safety, or compliance of information feeding into AI responses.
• Data exposure: Searching across the open internet failed SNCF's security requirements for a critical national infrastructure operator. Even a small query leakage could expose sensitive operational and strategic initiatives. This could result in consequences such as regulatory scrutiny, financial penalties, mandatory disclosures, and reputational damage.
• Scale uncertainty: With hundreds of thousands of employees, an intelligence layer needed to handle sustained, high-volume query load reliably. Further, with hundreds of thousands of employees, small risks can compound quickly. Even a single weak source or exposed query pattern could propagate across thousands of interactions, turning isolated issues into systemic vulnerabilities.

Why They Chose Linkup

Out of all the leading search providers, Linkup was the only one that could meet the demands of national workforce deployment where domain filtering, private network connectivity, and enterprise security controls are all non-negotiable.

Key advantages of Linkup:

Domain control at the API level

Linkup's /search endpoint enables domain whitelisting, the ability to restrict search results to a defined list of approved sources, and blacklisting, the ability to block specific domains entirely. This gives SNCF precise control over which sources employees can and cannot retrieve on every query. Read more about source filtering here.

Private cloud-to-cloud connectivity

Linkup delivers API access over a private network link rather than the public internet, keeping all search traffic within a controlled, encrypted infrastructure perimeter. This architecture eliminates the risk of query exposure.

Zero data retention and EU infrastructure

Linkup is the only European search provider and is designed for enterprise-grade compliance, including: SOC2 Type II certification, Zero Data Retention, and Data Processing Agreements.

Implementation

Linkup is embedded as the intelligence layer within SNCF’s internal GPT deployment.

All employee search queries are routed through Linkup’s /search endpoint, where configured domain rules are enforced before relevant context is returned to the platform. Connectivity between SNCF’s cloud environment and Linkup’s infrastructure is established through a dedicated private link.

Results

With Linkup as the intelligence layer powering their enterprise AI, SNCF was able to achieve:

• Web search accessible to 150,000 employees with highest security standard for query processing
• Complete domain control on every search
• Infrastructure that handles large-scale query volumes without rate-limiting or added latency
• EU data residency, Zero Data Retention, and DPA covered from day one

For any large enterprise or regulated organisation deploying AI assistants and platforms to a workforce at scale, the gap between a general-purpose search API and one built for enterprise requirements is exactly where Linkup fits.

Ready to give your AI systems the context they need? Try it for free here or learn more from our team.