Guest: Frank Karlitschek (LinkedIn)
Company: Nextcloud
Show Name: The Source
Topics: Open Source

As enterprises race to adopt generative AI, most rely on centralized models that require sending sensitive data to third-party clouds. Frank Karlitschek, Founder and CEO of Nextcloud, believes there’s a better way. In this TFiR clip, he explains how Nextcloud integrates AI locally — preserving user privacy while delivering the same functionality offered by commercial tools like Copilot and Gemini.

“Generative AI is a no-brainer for collaboration software,” Karlitschek said. “But integrating it the wrong way would mean giving companies like OpenAI access to all your data. That’s not what we want.”

Instead, Nextcloud’s engineering team built an AI framework that runs entirely within the user’s infrastructure. The system supports a range of open source models, including Falcon, Llama, Mistral, and DeepSeek, which power features such as live transcription, meeting summaries, translation, and intelligent email responses. By running everything locally, organizations can enjoy AI-driven productivity without sacrificing control or compliance.

Karlitschek also hinted at what’s next: the expansion of Nextcloud’s agentic AI features. The new assistant can automate tasks like scheduling meetings, summarizing calls, or sending follow-up messages — all while operating securely within the user’s private environment. “Your assistant knows everything about you,” he said, “but it never leaves your server.”

This privacy-first philosophy aligns perfectly with Nextcloud’s broader mission of decentralization and data sovereignty. As companies increasingly weigh the risks of cloud-hosted AI, Nextcloud is showing that innovation and privacy don’t have to be at odds.