The AI giants’ new problem: open AI - The Verge
Open-weight AI models can be poisoned for under $100. The frontier labs' strategy to stay closed is colliding with a diffused, distributed threat.
Why it matters
Open-weight model proliferation is creating a new security and capability liability for frontier labs. Poisoning attacks are cheap, easy, and reproducible—forcing a reckoning about the tradeoffs between openness and safety that will shape which labs win.
The key facts
5 to knowOpen-weight AI has diffused globally, research shows
Model poisoning attacks cost under $100 to execute
Poisoning described as 'laughably easy' by researchers
Frontier labs face strategic choice: closed vs. open model distribution
Safety/security implications of distributed open-weight ecosystems emerging as policy issue
Go to the source
Reuters Technologynews.google.com
Publisher excerpt: The AI giants’ new problem: open AI The Verge Nageswaran: Open weight AI has already diffused across the world—with dangers we must act upon livemint.com The Great Fracture Over Who Gets to Build the Future Lavender Hotel It’s Laughably Easy to Poison Open-Weight AI Models, Researcher Finds…