Project Lightwell, Taiwan’s AI Bet and the Open Model Safety Problem
The next phase of AI is being shaped not just by bigger infrastructure bets, but by growing pressure around software security, model safety and the real-world consequences of scale.
On this episode of Techstrong Gang, Mike Vizard, Jon Swartz, Fred Wilmot and Gina Rosenthal break down three stories that show where that pressure is building. The panel starts with Project Lightwell, a new $5 billion initiative from IBM and Red Hat that brings together lessons from frontier AI models and a 20,000-engineer network to help enterprises better secure open source software. The move signals how seriously major players are taking software supply chain security in the AI era.
The second segment turns to NVIDIA’s growing AI infrastructure push in Taiwan. Jensen Huang said the company plans to ramp annual spending there to around $150 billion, underscoring how central Taiwan has become to the future of chips, AI infrastructure and geopolitical competition.
The final segment, Open Source AI Oops, looks at reports that free software can strip guardrails from open AI models from major players like Meta and Google. The broader issue is whether open access is colliding with model safety in ways the industry still is not prepared to manage.
Taken together, these stories point to a broader shift: the AI boom is no longer just about capability. It is about securing the stack, managing the geopolitical footprint of infrastructure and deciding how much openness is safe in practice.
