Part 3/10:
Among the session's highlights were queries from attendees about deploying AI models securely and ensuring user trust. A recurring theme was understanding the fate of models after deployment, the handling of sensitive data, and maintaining transparency with app and infoSec teams. These questions reflect an industry eager to address the real-world complexities of AI deployment.
Addressing Security in AI: Skylow's Approach
A focal point of the discussion was how Skylow tackles the challenge of securing sensitive data in AI workflows. Roshik cited a recent MIT paper revealing that 95% of AI pilots fail to reach production, citing issues ranging from hallucinations and inaccurate outputs to security and scalability concerns.