CHANGELOG

What’s new?

A changelog of new features, technical improvements and enhancements lately

Feature

Red Team Attacker Mode

Test target models with a dynamic, multi-turn attacker that adapts its strategy based on each response.

The attacker generates scenarios from more than 170 vulnerability seeds and continues the interaction according to configurable turn limits. Potentially successful attacks are validated by a separate judge model, while complete prompts and responses can be exported as CSV or Excel files for review, reporting, and sharing.

Enhancement

Async Reasoning Early Exit

Return security decisions faster when any parallel check identifies a definitive violation.

The reasoning engine stops waiting for unfinished checks once a request is confirmed as unsafe, reducing latency and unnecessary compute. Safe requests still complete the full evaluation process, preserving the existing level of security coverage.

Enhancement

Customizable Error Messages

Customize the error messages returned when a security policy is violated to match organizational tone and communication standards.

Each response now includes a Log ID for faster investigation, while BG header support allows different applications or integration points to receive tailored messages under the same policy.

Feature

Secondary Model Routing

Automatically route requests to a secondary model when the primary model is unavailable due to timeouts, service outages, provider errors, or rate limits.

When PII or PHI is detected, requests can also be redirected to a local model so sensitive data remains within the organization. Routing is handled automatically through the primary and secondary model configuration in the Model Library.

Enhancement

Multiple Endpoint Tokens

Create and manage multiple independent access tokens for a single endpoint.

Separate tokens can be assigned to different applications, integrations, or environments such as development, testing, and production. Each token can be tracked, rotated, or revoked independently without interrupting other clients using the same endpoint.

Integration

LDAP, AD, and SSO Integration

Connect BeyondGuard directly to existing enterprise identity and access management systems.

Users can sign in with their corporate accounts through LDAP, Active Directory, or SSO. Organization information is transferred automatically during authentication, ensuring that each user can access only the relevant policies, projects, models, and logs.

Integration

ICAP Integration

Inspect enterprise proxy and gateway traffic with faster, more flexible ICAP connectivity.

Expanded adapter support, real-time request visibility, and asynchronous processing reduce latency and allow multiple requests to be inspected in parallel, even under heavy traffic.

Enhancement

GPU Load Balancing

Maintain reliable AI security performance through GPU-based load balancing and automatic service scaling.

Incoming requests are routed to the active server with the lowest load, while unavailable servers are automatically bypassed. Security services scale according to demand, supporting high traffic volumes without interrupting availability or slowing down inspections.

Feature

Expanded Red Team Coverage

Evaluate AI systems with 50,000 built-in test cases across prompt injection, dangerous code, toxicity, PII and secrets exposure, and HTML script injection.

The release also adds multi-turn injection checks and application-level Red Team testing, extending coverage beyond the model to business logic, APIs, authorization flows, tool calls, and chained attack scenarios.

Enhancement

Centralized Model Library

Manage local, self-hosted, and cloud-based AI models through a single central inventory.

Model connection details and credentials are stored securely and reused across modules such as Guard, Sandbox, Red Team, and Agent Builder. Available models can also be fetched directly from providers, so updates made in the Model Library are reflected across the platform.

Feature

Shadow AI Detection

Detect unauthorized use of third-party AI services across the organization through network traffic analysis.

Shadow AI can analyze traffic from HTTP mirrors, Kafka, or database sources, with configurable detection sensitivity. Identified activity is logged with provider, confidence score, request details, and timestamps, then filtered, reviewed, or exported as CSV.

Enhancement

Observable Model Sandbox

Analyze externally sourced AI models in an isolated environment before moving them into production.

Models from Hugging Face, Git, or custom sources can be scanned for malware, file integrity issues, and risky scripts or code. Each stage of the scan can be monitored in real time, with detailed events, results, and a downloadable report available from a single screen.

Feature

Multi-Policy Management

Create and manage separate security policies for different organizations, teams, and user groups.

Policies can be versioned, prioritized, copied, and assigned directly to endpoints. This allows different business units to operate with independent rules and risk settings on the same platform.