SECURITY

Beyond Identity launches Ceros AI agent security platform

Beyond Identity Inc. today debuted a platform called Ceros that is designed to help enterprises secure their artificial intelligence agents.

New York-based Beyond Security is backed by more than $200 million in funding. Its flagship product is an IAM, or identity and access management, platform that companies use to regulate employee access to their applications.

The software can enforce multifactor authentication and block login requests from insecure devices. Ceros extends those cybersecurity controls beyond employees to AI agents.

A large enterprise can have upwards of hundreds of AI automation workflows. As a result, tracing a suspicious file change to the specific AI agent that carried it out is often challenging. Beyond Identity says that Ceros automates the process. The platform logs every AI agent session complete with data on which user launched the session and from what device. 

The platform also monitors the external components on which AI agents rely to work. Ceros can map out all the cloud-hosted large language models, MCP servers and local programs used in an automation initiative. Administrators can define controls to regulate which agent may use what tool and under what circumstances.

Ceros-managed agents access external services using a credential called a device-bound passkey. It can’t be moved off the system on which an agent runs, which means that hackers have no way of stealing it. Furthermore, the platform mitigates prompt injection attacks and attempts to access AI workloads’ application programming interface keys. 

Beyond Identity says its platform also eases certain other tasks. When the LLM that powers an AI agent goes offline, developers can avoid downtime by rerouting user prompts to a different model. Building an automated LLM switching mechanism usually requires a significant amount of work. Ceros automates the task, which enables developers to focus on other aspects of their AI agent project. 

‘Today’s AI threats demand strong identity and device trust at the point of access, before a risky agent, user or device ever hits your network,” said Beyond Identity Chief Executive Officer Jasson Casey. ‘With Ceros, CISOs get complete visibility and control across every agent, identity, device and action with a platform built for autonomous AI at scale.”

Several startups offer competing products that can likewise map out the AI agents in a corporate network. One of those startups, Cyera Ltd., raised $600 million in funding at a $12 billion valuation last week. The company sells its AI observability features as part of a platform that can also help companies find sensitive records in their internal datasets.

The strong industry interest in AI agent access controls may prompt OpenAI Group PBC and Anthropic PBC to integrate more such capabilities into their respective products. The two companies both maintain a fast product release cadence. In February, OpenAI introduced an AI agent management platform that includes cybersecurity controls.

Image: Unsplash