The AI Native Dev - from Copilot today to AI Native Software Development tomorrow How DeepSeek leveraged Qwen and Llama to build its model in $5M
Apr 7, 2026
Amanda Brock, CEO of OpenUK and open technology law expert, discusses true openness in AI and the rise of openwashing. She covers how DeepSeek built a frontier model for $5M using Qwen and Llama. She compares Western and Chinese open-source approaches and explains how to verify model openness and why national strategies and community-building matter.
AI Snips
Chapters
Transcript
Episode notes
Evaluate Openness By Component
- Open AI should be disaggregated into components so openness is evaluated per part not as a single claim.
- Amanda Brock recommends judging model, algorithm, dataset, and agent individually and checking each against OSD-compliant licenses like GPL or Apache.
Openness Drives Fast Innovation
- Open models accelerate innovation and democratize AI access, especially for countries outside US and China.
- Brock notes Llama's opening spurred unprecedented pace of work and enabled derivative projects like DeepSeek's R1.
Open Washing Destroys Trust
- 'Open washing' erodes trust because projects may claim openness while keeping commercialization or acceptable-use restrictions.
- Brock argues Meta's Llama had restrictions (commercial triggers and acceptable use) that broke the core open-source guarantee of 'use for any purpose.'