Cisco AI Defense

AI security software for discovering enterprise AI assets, scanning supply chains, validating models and agents, and applying runtime protection across connected environments.

About Cisco AI Defense

Cisco AI Defense helps enterprises secure the AI applications they develop and the AI services their employees use. Its capabilities cover asset discovery, supply chain risk detection, model and application validation, runtime protection, and access controls. The platform addresses security, privacy, and safety risks across the AI lifecycle.

For development teams, AI Defense combines scanning with algorithmic red teaming to identify weaknesses before release. Production controls inspect AI interactions and apply guardrails against threats such as prompt injection and sensitive-data exposure. Current documentation also covers agent behavior and MCP requests and responses.

Cisco describes deployment coverage across cloud, VPC, and on-premises AI POD environments. Available capabilities depend on the purchased package and associated deployment requirements.

Key Features

AI asset visibility: Discovers models, applications, agents, and connected resources across distributed environments.
Supply chain risk management: Scans AI components, repositories, and MCP servers for malicious or unsafe content.
Model and application validation: Uses algorithmic red teaming to evaluate security and safety weaknesses.
Runtime guardrails: Inspects AI interactions and applies controls against adversarial inputs, harmful responses, and data leakage.
Agent and MCP protection: Addresses risks involving tool misuse, memory poisoning, privilege escalation, and agent workflows.
AI access controls: Supports policies governing employee access to third-party AI applications through the required Cisco Secure Access package.

Pricing

Subscription options: Application-based subscriptions or committed-usage purchasing. Public monetary rates are not specified.
Application-based packages: The offer document names Advantage, Validation Essentials, and Runtime Essentials. Applicable runtime terms assume 10 million queries per AI application annually. Advantage and Runtime Essentials include 3,504 Multicloud Defense Premier gateway hours per AI application annually.
Committed-usage blocks: Runtime Guardrails use blocks of one billion inspection tokens; supply chain management uses 50 governed assets; validation starts with 600 tests, followed by blocks of 120 tests. Full blocks are deducted upon initial use, without partial refunds.
Additional consumption: Excess usage can require additional entitlements or incur overage charges. Published monetary overage rates were not established.
Support and dependencies: Enhanced Support is included; Premium Support is optional. AI Access requires Cisco Secure Access Advantage.
Subscription changes: Upgrades and additional committed usage are permitted during the term; reductions to the subscription tier or committed usage are not.

Use Cases

Identify enterprise AI applications and their associated resources.
Scan models and MCP components before integrating them into applications.
Validate AI models and agents before production release.
Apply runtime protection to AI interactions across supported infrastructure.

Pros & Cons

Pros:

Combines component scanning, behavioral testing, and runtime protection.
Documents security controls for agent workflows and MCP interactions.
Supports deployment scenarios spanning cloud and on-premises AI infrastructure.

Cons:

AI Access depends on Cisco Secure Access Advantage.
Committed-usage blocks are indivisible, and subscription commitments cannot be reduced during the term.

Integrations

Cisco Secure Access, Cisco Multicloud Defense

FAQ

What does Cisco AI Defense protect?

It provides security capabilities for AI models, applications, agents, connected components, and enterprise use of third-party AI services.

Does Cisco AI Defense include red teaming?

Yes. Its model and application validation capability uses algorithmic red teaming to examine security, privacy, and safety weaknesses.

Can Cisco AI Defense secure MCP interactions?

Yes. Current documentation covers MCP supply chain scanning and runtime inspection of MCP requests, responses, and agent interactions.

Does Cisco AI Defense support on-premises deployment?

Cisco describes coverage across cloud, VPC, and on-premises AI POD deployments. The architecture depends on the selected offering.

What is Cisco AI Defense Explorer?

Explorer is an offering for AI builders to red team applications before moving to an enterprise deployment. Its exact access allowances were not verified.