Cohesity unveils Enterprise AI Resilience for safe scale

Cohesity has launched an Enterprise AI Resilience strategy focused on protecting AI infrastructure, governing sensitive data used by AI systems, and enabling rapid recovery from unintended or malicious actions by AI agents.

It is positioning the strategy as a unified approach for organisations moving beyond AI pilots and putting AI into production. Cohesity argues this shift changes operational risk, as AI agents and automated workflows interact directly with business systems and data.

Broader risk

Agent-based AI systems introduce new points of failure for IT and security teams. Logic errors, corrupted inputs, and prompt manipulation can trigger incorrect actions across connected systems. These can occur at machine speed and cause wider disruption than conventional software faults.

The approach covers protection for components around AI models, including vector databases, agent memory, model configuration and policy layers, and training and fine-tuning datasets. It also extends to enterprise data stores that feed AI applications.

Cohesity uses immutable snapshots of AI environments and synchronised, point-in-time recovery for agents, data, and supporting infrastructure. It says it can restore files, databases, object storage, SaaS applications, vector stores, and agent memory, aiming to reduce downtime without full rebuilds.

Agent containment

A central element is recovery from rogue or unintended agent behaviour. Cohesity argues detection alone does not address the speed of automated mistakes, and calls for coordinated containment and restoration across the enterprise estate, including IT service management and observability tools.

Integrations with ServiceNow and Datadog form part of the workflow design. Cohesity says it can translate risk signals into automated recovery actions through API-driven workflows, with anomalous behaviour and policy violations triggering point-in-time restoration.

"By strengthening defence and enabling secure data activation, Cohesity is establishing enterprise AI resilience as the foundation for responsible, high-velocity AI adoption," said Sanjay Poonen, CEO and President of Cohesity. "Enterprises need the confidence to manage AI-driven risk and recover quickly when disruptions occur. Cohesity provides the resilience foundation that protects AI infrastructure, governs data access, mitigates agent-driven risk, and unlocks the transformative power of trusted enterprise data."

Data governance

The strategy also includes data governance for AI environments across cloud, SaaS, and hybrid deployments. Cohesity says continuous visibility and controls are increasingly important as AI systems draw from broader pools of enterprise information.

Cohesity Data Security Posture Management is part of the offer. Powered by Cyera, the product discovers and classifies sensitive data, monitors access patterns, and supports governance controls. Cohesity also links governance with recovery, aiming to restore affected data and AI systems to trusted states after exposure or misuse.

"AI is transforming enterprise operations, but it also introduces new complexities across infrastructure, data governance, and agent-driven automation," said Greg Statton, VP and Chief Technology Officer APJ at Cohesity. "By advancing Cohesity's Enterprise AI Resilience architecture, we're delivering deeper threat visibility, stronger control over data and agent behaviour, and faster recovery when signals indicate risk. At the same time, we're empowering organisations to activate trusted enterprise data securely for AI and analytics. Together, enhancements such as integrated malware scanning, coordinated agent and cloud recovery, and expanded sovereign cloud partnerships provide the resilience needed to adopt and scale AI with confidence."

Cohesity is also targeting regulated requirements in Australia and New Zealand, emphasising compliance controls for data access. "Across Australia and New Zealand, customers want to harness AI to drive efficiency and growth, but only if they can do so securely with minimal risks," said James Eagleton, Managing Director, ANZ, Cohesity. "By strengthening protection across data, agents, and cloud environments, Cohesity is enabling ANZ businesses to confidently operationalise AI while meeting their security and compliance requirements."

Data activation

Alongside defensive controls, Cohesity is adding features it describes as data activation for AI and analytics. It says organisations hold large volumes of unstructured and time-series data in the Cohesity Data Cloud, and that governed access can make that information usable for AI tools.

Cohesity has announced federated semantic search via the Model Context Protocol. It says this allows AI-powered enterprise tools to access governed backup data without duplicating it, citing Glean as an example.

"AI agents are only as useful as the information they can securely access," said Zubin Irani, VP of Partnerships at Glean. "Enterprises have decades of critical knowledge stored across systems, but much of it remains locked away. By enabling federated access to governed data in the Cohesity Data Cloud, we're helping organisations bring that trusted context into Glean so AI can deliver more accurate answers and actions while maintaining strong security and compliance controls."

Cohesity also outlined plans for an upcoming Gaia Catalog, which it says extends its Gaia AI platform. The company says the catalogue will provide access to protected data from analytics platforms such as Databricks and Microsoft Fabric, without duplicating data or rebuilding ETL pipelines.

The strategy is delivered through the Cohesity Data Cloud, covering protection for on-premises, cloud, and SaaS environments. Cohesity says it links data protection, governance, and recovery with access controls for AI and analytics use cases.