Wire · regulatory
'Open-Washing' Is Everywhere in AI. Four Criteria Cut Through It
◆ Sectors
◆ Geography
◆ Source
◆ Verified
Fusion42 · 25 July 2026 · Fusion42 review
An AI policy analyst applies rigorous open-source criteria to assess whether models like Moonshot's Kimi K3 genuinely qualify as open-source, finding widespread 'open-washing' where companies claim openness while withholding training data, code, and transparent governance. The Open Source Initiative's formal definition requires disclosed training code, weights, data, and unrestricted licensing; very few models meet this bar.
This Wire brief sits within Fusion42's coverage of AI Frontier Models and Generative AI.
◆ ◆ The Wire takeaway
If you're building on top of an 'open-source' model, check whether you actually have the training data and pipeline—most don't, which means you're building on a closed platform that can change its terms or behaviour without warning. The four-criterion test (architecture, code, weights, data all public under OSI licence) eliminates the noise; OLMo, GPT-NeoX, and K2 pass; Kimi K3 doesn't.
◆ Coverage
1 source · 25 Jul 2026
◆ Related on Wire
◆ Topics