DeepKeep Partners with TrueFoundry to Bring Runtime AI Security to the AI Gateway

DeepKeep is partnering with TrueFoundry to integrate DeepKeep’s AI Firewall with the TrueFoundry AI Gateway.

The integration gives organizations a centralized way to apply runtime AI security across the models and applications already routed through TrueFoundry.

TrueFoundry provides the control layer for enterprise AI traffic. DeepKeep provides the security layer that evaluates live AI interactions and returns an enforceable decision: allow, block, or redact.

Bringing AI Security Into the Gateway Path

Enterprise AI environments rarely stay confined to one model or one application.

Different teams use different providers. Models change. New applications reach production. Security requirements, meanwhile, are expected to remain consistent across all of them.

Implementing separate security logic inside every application creates unnecessary engineering work and uneven enforcement. One application may receive the latest policy while another continues running older controls. A newly introduced model may require another integration project before security teams can apply the same protections.

An AI gateway creates a shared control point.

The TrueFoundry AI Gateway provides a unified layer for model access, routing, fallback, observability, rate limiting, and policy management. Applications send AI traffic through the gateway rather than maintaining separate connections and operational controls for each model provider.

The DeepKeep integration adds runtime AI security to that shared path.

How the Integration Works

Organizations configure DeepKeep’s AI Firewall as a security integration within the TrueFoundry AI Gateway.

When covered AI traffic passes through the gateway, the interaction can be evaluated against the security policies configured in DeepKeep. DeepKeep then returns one of three decisions:

Allow

Interactions that comply with the organization’s policies continue through the gateway.

This keeps legitimate traffic moving without requiring developers to implement security logic separately within every application.

Block

Interactions that violate an enforced policy are stopped.

This can include activity such as prompt-injection attempts, jailbreaks, sensitive-data exposure, or other behavior that falls outside the organization’s permitted use of AI.

Redact

Sensitive or restricted content can be removed while the permitted remainder of the interaction continues.

Redaction provides a more precise response than blocking the entire request when the underlying use case is legitimate but some of the content should not be sent to or returned from an AI system.

TrueFoundry’s guardrail architecture supports both validation and mutation of AI traffic. Validation allows a security service to approve or reject an interaction. Mutation allows approved content to be transformed before it continues, which enables controls such as redaction.

Clear Responsibilities Across the AI Stack

The integration separates two related but distinct responsibilities.

TrueFoundry manages how AI traffic reaches models. It provides the operational control plane for routing, provider access, reliability, usage, and gateway-level policies.

DeepKeep determines whether the AI interaction complies with the organization’s security requirements.

This separation lets platform teams continue managing AI infrastructure through TrueFoundry while security teams apply DeepKeep policies to the same production traffic.

Applications do not need to become their own security enforcement systems. Developers can continue using the gateway path already established by the organization.

Consistent Protection Across Models and Applications

The main advantage is consistency.

Security policy can be applied at the point where AI traffic converges instead of being recreated across every application and model integration.

When teams introduce another application or route traffic to a different model, the security layer does not have to be reinvented. The relevant gateway rules can continue directing covered traffic through DeepKeep’s AI Firewall.

This creates a cleaner operating model:

• Platform teams manage models, routing, and gateway availability.
• Security teams define and tune runtime AI policies.
• Application teams use the organization’s approved gateway path.
• Policy decisions are enforced while interactions are occurring.

The result is a practical way to scale AI adoption while retaining centralized security control.

Getting Started

The DeepKeep integration with TrueFoundry AI Gateway is available through TrueFoundry’s guardrail configuration.

Technical setup instructions and integration details are available here: www.truefoundry.com/docs/ai-gateway/deepkeep

‍