As AI agents go mainstream, companies lean into confidential computing for data security
Briefly

Companies must address data security as AI agents manage internal data movement. Confidential computing offers a protective hardware boundary that ensures AI models and data remain secure. Regulations demanding auditability in sectors like healthcare and finance necessitate oversight of AI models. Google is allowing in-house operations of its proprietary Gemini models using confidential computing technology on Nvidia GPUs, enabling companies to run Gemini without relying on internet or Google Cloud connectivity. This aligns with enterprises' desire for control in private-cloud AI environments.
"Confidential computing creates a hardware boundary in which AI models and data are locked. Information is released only to those models and agents with proper access to prevent unauthorized use of protected data."
"For enterprises concerned about AI security, putting confidential computing into action can alleviate some of those fears. It goes hand-in-hand with enterprises' preference for private-cloud AI."
Read at Computerworld
[
|
]