The article reports that many companies are adopting open or partially open AI models despite security and governance concerns, because they offer lower cost, greater customization, and more control than proprietary systems. Firms say smaller, adaptable models are often better suited for business-specific tasks, and most organizations use a mix of open and closed models depending on the use case. The article concludes that while open models introduce risks such as supply-chain vulnerabilities and potential backdoors, companies believe the flexibility and performance benefits make those risks manageable.