DICE

A new analysis examines the cybersecurity trade-offs of open-source and open-weight AI models, arguing that freely available model weights carry both benefits and risks. Openness enables scrutiny, transparency and independent auditing, but also makes it easier for malicious actors to fine-tune or abuse powerful systems. The piece frames adoption as a calculated gamble that organisations must weigh carefully.