Shivaami and Google Cloud tackle multi-agent enterprise security

As enterprises move from individual AI experiments to multiple intelligent agents, the bigger challenge is becoming clear. Shivaami and Google Cloud put data access, security guardrails and real-world deployment at the centre of that shift.

author-image
DQChannels Bureau
New Update
Shivaami and Google Cloud tackle multi-agent enterprise security

AI agents may be getting smarter, but scaling them across an enterprise creates a different problem. They need access to data spread across CRM platforms, finance systems and internal documents, while security teams still need control over what those agents can access.

Advertisment

That tension was at the centre of a session hosted by Shivaami in collaboration with Google Cloud at the Google Cambridge office. The event brought together technology, security and AI leaders to discuss how organisations can build a Multi-agent enterprise without allowing governance to fall behind.

Multi-agent AI runs into a data problem

One of the key discussions focused on data access. AI agents can only work with the information available to them, but enterprise data often sits across disconnected systems.

The session highlighted the need to bridge CRMs, finance systems and internal documents into a more cohesive view. For organisations, this makes data connectivity an important part of scaling Enterprise AI, rather than treating agents as isolated tools.

Advertisment

Cloud security needs to keep pace with agents

The second challenge is security. As more agents are deployed, organisations risk creating what the session described as shadow AI, where systems operate without enough oversight.

The event focused on active, real-time guardrails that can prevent agents from accessing sensitive information without authorisation. This shifts the security conversation from simply deploying agents to controlling what they can do as they operate.

Shivaami Google Cloud session puts AI into practice

The practical side of the discussion came through a live Gemini demonstration. A single agent pulled context from disconnected systems before automated guardrails immediately stopped an unauthorised access attempt.

Advertisment

The event also explored how Gemini Enterprise could operate alongside self-hosted, open-weight GPU models across different IDEs and agent frameworks. The discussion with Mike Hiltz, VP and CISO at nference, brought the focus back to real-world execution across varied technology environments.

Read More: 

TAG national distributor deal expands Apacer’s India reach

Kaspersky exposes iPhone 18 scams using fake deals

Fortinet joins Ingram Micro's Private Offers on AWS Marketplace

Aiven India launch puts AI-Ready data infrastructure first