SK Telecom Invests $100M in Anthropic to Fine-Tune Claude
South Korean telecom giant SK Telecom has invested $100 million in Anthropic to co-develop a customized, multilingual version of the Claude LLM tailored specifically for the global telecommunications industry.

SK Telecom (SKT), the largest mobile operator in Korea, has made a strategic investment of $100 million in Anthropic. This funding, which builds upon a previous investment from SK Telecom Venture Capital, accompanies a new commercial partnership aimed at building a customized large language model specifically optimized for the telecommunications industry.
The collaboration will leverage Anthropic's Claude model, using fine-tuning techniques to adapt the system to telco-specific needs. SKT's industry experts will provide direct feedback on Claude's responses to train the model on specialized solutions. The resulting multilingual model will support languages including Korean, English, Japanese, and Spanish. It is designed to power a variety of industry applications, such as customer service, marketing, sales, and interactive consumer tools.
For AI practitioners and developers in the telecommunications sector, this partnership signals the arrival of highly specialized, domain-specific foundation models. Instead of relying on generic LLMs that require extensive prompt engineering or independent fine-tuning, developers will gain access to a model pre-trained on telco customer service and marketing data. This should significantly reduce deployment times and improve accuracy for customer-facing applications.
Alongside the SKT partnership, Anthropic announced a $5 million grant program to fund independent research into how AI impacts user wellbeing. The company also rolled out updates to its Claude Fable 5 model, modifying its biology safeguards to reduce false positives. This update aims to decrease the frequency of fallbacks, where the system unnecessarily switches to a less capable model during biology-related queries.
This is our own summary of reporting by Anthropic



