Data security platform company Veza Inc. today announced the launch of AI Agent Security, a purpose-built product that helps organizations secure and govern artificial intelligence agents at enterprise scale.

AI Agent Security, which is built on Veza’s Access Graph, offers unified visibility into AI agents across leading platforms and allows security teams to answer critical questions such as “What AI agents exist in our environment?” and “Which humans can control or act through AI agents?”

The new service seeks to assist businesses as they accelerate AI adoption by giving security and governance teams the visibility and control needed to protect data and enforce trust across human-AI interactions.

AI Agent Security has also been designed to help businesses deal with the rise of what Veza describes as a “new class of security risks,” such as prompt injection attacks, in which outsiders can manipulate agents into revealing confidential data or executing unauthorized actions. Veza argues that these types of attacks depend on AI agents having access to read data input by an attacker, commonly by sending emails, calendar invites or website sales forms.

Given the rising risks, full visibility and data-driven insights into the scale and scope of AI agents are needed and this is where AI Agent Security steps in.

AI Agent Security eliminates AI blindspots by discovering and classifying agent identities across the enterprise by providing a single platform to visualize how the AI agents are connected with data resources and large language models. The service enforces the principle of least privilege by laying out the full blast radius of agents and highlighting what sensitive data and system resources they can access.

Other features include the ability to establish robust AI governance with complete agents-to-human identity mapping that allows enterprises to govern which users and groups can deploy, manage and interact with AI agents.

The service also ensures continuous compliance with support for Open Worldwide Application Security Project recommendations for user/machine access audits and provides instant answers to what the agents have access to for auditors and enterprises to meet their regulatory requirements.

AI Agent Security integrates with major AI ecosystems, including Microsoft Copilot Studio, Salesforce Agentforce, AWS Bedrock and Google Vertex AI and supports access governance for OpenAI Group PBC and Azure AI services.

“It’s the Wild West with agentic AI right now,” said co-founder and Chief Executive Tarun Thakur. “Chief information security officers and chief information officers are racing to deploy AI, but few have visibility or control over what these agents can actually do.”

Thakur contends that “Veza AI Agent Security brings order to that chaos by unifying discovery, governance and control for AI systems. In the end, all roads lead to identity: understanding who or what has access, and why.”

The new release may be one of the final ones Veza makes as an independent company. last week cloud software maker ServiceNow Inc. signed an agreement to acquire Veza for a figure believed to be more than $1 billion.

Image: SiliconANGLE/Ideogram