Ceros has launched what it describes as the industry’s first trust layer built specifically for autonomous AI agents. As enterprises rapidly adopt agentic AI, many security teams lack visibility into their AI attack surface, including exposed data, credentials, and high-risk tools. Ceros is designed to address these challenges by providing security and governance capabilities tailored to the operational realities of autonomous systems.
From agents acting independently leveraging local tools or MCP ecosystems, Ceros enables enterprises to unpause stalled AI initiatives, contain shadow AI and safely accelerate enterprise AI adoption by providing a control layer that enables AI adoption securely.
“Ceros addresses agent security as an operational problem, not just a compliance checkbox. It monitors potential data exfiltration, enforce controls around sensitive tool and system access, and reduce real-world agentic engineering risks like prompt injection, over-permissioned workflows, and unintended actions across local and connected environments,” said Neal Mueller, Product at Apple.
Ceros anchors provable security via a five-dimensional identity vector, including:
- Who is running the agent?
- How was the user behind the agent authenticated and authorized?
- Is the device the agent is running compliant with the organization’s policies?
- What applications, tools or skills is the agent running?
- What agent configuration it launched with?
“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. With Ceros, CISOs get complete visibility and control across every agent, identity, device and action with a platform built for autonomous AI at scale,” said Jasson Casey, CEO of Beyond Identity, the developer of Ceros.
AI Discovery & Inventory: Continuously discover and inventory the entire AI ecosystem: AI agents, tools, MCP services, providers, and models operating across enterprise environments.
- Detect installed and running agents
- Discover MCP tools, built-in tools, and remote MCP services
- Identify AI providers and inference engines in use
- Track model usage and runtime activity
- Sanction / Unsanctioned AI assets
- Catalog-driven discovery with extensible customer-defined signatures
Runtime Governance & Policy Enforcement: Govern agent execution through continuously evaluated authorization policies and enforceable runtime controls.
- Define authorization envelopes for agent sessions
- Control which agents may launch
- Restrict tool access and execution scope
- Enforce argument-level policy controls
- Continuously evaluate runtime behavior
- Alert, degrade, or terminate violating sessions
- Prevent agents from pushing directly to production systems.
AI Provenance, Audit & Incident Response: Maintain complete provenance and forensic visibility across agent sessions, conversations, tools, and actions.
- Log conversations and execution metadata
- Track agent, user, and device provenance
- Attribute actions to specific sessions
- Support security investigations and compliance workflows
- Maintain inspectable execution histories
AI Runtime Orchestration & Resilience: Inject, replace, and manage tools and AI providers dynamically to improve resilience, compliance, and operational consistency.
- Inject approved tools into agent environments
- Replace risky or non-compliant tooling
- Redirect agents to alternate LLM providers
- Maintain continuity during provider outages
- Standardize enterprise-approved AI infrastructure
To learn more about Ceros, please visit: https://ceros.sh
Related News:
HP Research: Legitimate Remote Access Tools Used as Backdoors
Netwrix Data and Identity Report Reveals Gaps in AI Security Readiness
