Protect and evaluate AI applications, models, agents and Model Context Protocol deployments with TrojAI Defend, TrojAI Detect and Defend for MCP, including customer-controlled cloud or on-premises deployment options.
Technical detail
Core features
Runtime prompt-injection, jailbreak and data-leak protection
AI model security testing and evaluation
Agent and Model Context Protocol discovery and control
Customer-controlled cloud or on-premises deployment
Mappings to OWASP, MITRE and NIST AI-security guidance
Agent-action, MCP invocation and tool-call telemetry alongside coding-agent protection
AI security
Prompt injection
Model testing
AI agents
Model Context Protocol
Use & integration
Recorded applications
Securing generative-AI applications
Testing models before deployment
Governing enterprise AI agents and MCP connections
Protecting sensitive AI workflows in controlled environments
These describe the reviewed application scope, not confirmed deployments, released requirements or procurement eligibility.
Evidence of maturity
TrojAI is a current A10 AI-security offering after the completed June 2026 acquisition. A10 announced AI Gateway general availability on August 13 and describes it as complementing TrojAI across the AI lifecycle; independent deployment-specific performance remains undisclosed.
Commercial availability
TrojAI is offered through A10/TrojAI sales and demonstration routes, with customer-controlled deployment options. Product packaging, supported integrations and Canadian contracting arrangements are agreed per engagement.
Integration and verification
Readiness
A technology readiness level is not established by the reviewed public sources.
AI model, agent and application security could help protect autonomous monitoring software that uses those components, but no public source reviewed here establishes a specific patrol-system integration or deployment.
This is a reviewed public-source assessment. It is not procurement eligibility, endorsement, customer interest, or classified demand.
Original sources 4 linked records
These records support this capability and its published Mission area and Defence need connections. A source count is not a count of independent confirmations.
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technicaldomainslugs
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technical tags
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technicaldomainslugs
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technicaldomainslugs
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · core features
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · defence applications
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · defence applications
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · core features
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · core features
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technology readiness level
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · summary
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · core features
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · name
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technical tags
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · summary
The official capability material supports the named AI assurance and model security offering and the bounded operational description prepared for reviewer consideration.
Supports: Capability · core features
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technicaldomainslugs
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · commercial availability
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technical tags
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · core features
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · technical tags
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · defence applications
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · capability type
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Capability · defence applications
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Mission area · Autonomous Patrol and Monitoring
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Mission area · Autonomous Patrol and Monitoring
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Mission area · Autonomous Patrol and Monitoring
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
Supports: Mission area · Autonomous Patrol and Monitoring
Runtime prompt-injection and data-leak controls extend to agent actions, MCP servers and tools, with invocation/span/tool-call telemetry and pre-deployment evaluation. The current offer separates agent runtime protection, coding-agent security, red teaming and model guardrails; vendor efficacy claims require deployment-specific validation.
A10's current TrojAI catalogue supports runtime protection, model security testing, agent and MCP discovery or control, customer-controlled deployment and common AI-security framework mappings.
Supports: Capability · defence applications
A10's current TrojAI catalogue supports runtime protection, model security testing, agent and MCP discovery or control, customer-controlled deployment and common AI-security framework mappings.
Supports: Capability · name
A10's current TrojAI catalogue supports runtime protection, model security testing, agent and MCP discovery or control, customer-controlled deployment and common AI-security framework mappings.
A10 announced a completed acquisition of TrojAI on June 15 2026. A10 distinguished build-time AI red teaming from runtime threat protection and said portfolio integration was expected following acquisition, not already complete in that announcement.
A10 announced AI Gateway general availability on August 13 2026, describing it as complementing TrojAI and ThreatX across testing, runtime and network/API security. Software or integrated hardware runs in customer-controlled on-premises/private-cloud/air-gapped environments; this does not prove every product is a single integrated configuration.
Supports: Capability · maturity
A10 announced AI Gateway general availability on August 13 2026, describing it as complementing TrojAI and ThreatX across testing, runtime and network/API security. Software or integrated hardware runs in customer-controlled on-premises/private-cloud/air-gapped environments; this does not prove every product is a single integrated configuration.
Public sources cited · Facts and assessments kept separate · Human review
A10's current TrojAI catalogue supports runtime protection, model security testing, agent and MCP discovery or control, customer-controlled deployment and common AI-security framework mappings.
Supports: Capability · commercial availability
A10's current TrojAI catalogue supports runtime protection, model security testing, agent and MCP discovery or control, customer-controlled deployment and common AI-security framework mappings.
Supports: Capability · maturity
A10's current TrojAI catalogue supports runtime protection, model security testing, agent and MCP discovery or control, customer-controlled deployment and common AI-security framework mappings.
Supports: Capability · capability type
A10's current TrojAI catalogue supports runtime protection, model security testing, agent and MCP discovery or control, customer-controlled deployment and common AI-security framework mappings.
Supports: Capability · summary
A10's current TrojAI catalogue supports runtime protection, model security testing, agent and MCP discovery or control, customer-controlled deployment and common AI-security framework mappings.
Supports: Capability · technicaldomainslugs
True North Map classifies the current product under communications and cyber and mission software and data. The retained autonomous-patrol-and-monitoring link is rewritten as a needs-review software-security assessment, not a deployment or adoption claim.
Supports: Mission area · Autonomous Patrol and Monitoring
True North Map classifies the current product under communications and cyber and mission software and data. The retained autonomous-patrol-and-monitoring link is rewritten as a needs-review software-security assessment, not a deployment or adoption claim.
Supports: Mission area · Autonomous Patrol and Monitoring
True North Map classifies the current product under communications and cyber and mission software and data. The retained autonomous-patrol-and-monitoring link is rewritten as a needs-review software-security assessment, not a deployment or adoption claim.
