The fact that we don't know what it does is the feature
27 October 2026 (diginomica). Kelsey Hightower, open source expert and former Google Cloud engineer, discussing the unpredictability of AI models at a Civo Navigate panel in London in 2026.
How to manage and maintain open source models and contributions in the AI age
AI and safety · AI and philosophy · 2026
Who else?
What it refers to: AI models' unpredictable behavior, unlike traditional software such as Linux kernels or web APIs with known, fixed functions.
Other voices
- “None of us know anything at all” Ethan Mollick, Researcher,
- “I think of all agents as insider risk” Satya Nadella, Builder,
- “Making autonomous AI systems with a mandate to neutralize threats is incredibly dangerous” David Krueger,
- “No model is perfect, and anyone who claims theirs is shouldn't be trusted with public infrastructure” Hantz Févry,
- “I'm willing to accept no incidents of cybersecurity” Jensen Huang, Builder,










