What it is
Alice (formerly ActiveFence) is an enterprise security, safety, and trust platform targeted at organizations that build or operate generative AI applications, agents, foundation models, and user-generated content systems. It is presented as a suite for testing, protecting, and monitoring AI systems across development and production stages.
The platform is organized under a WonderSuite label that includes distinct capabilities: WonderBuild for pre-launch stress testing, WonderFence for dynamic runtime guardrails, and WonderCheck for ongoing automated red-teaming and drift detection.
The core AI capability is an adversarial intelligence engine called Rabbit Hole. Alice states Rabbit Hole is built on billions of toxic and abusive samples combined with cross-cultural expert analysis to train classifiers and support red-teaming, detection, and runtime protections.
Alice reports global scale metrics including coverage of 3B+ users, support for 120+ languages, and over 1B daily AI-human interactions, and states it safeguards more than 50% of online experiences.
Key features
WonderBuild provides pre-launch stress testing that simulates adversarial inputs and abuse patterns to evaluate model, application, and agent resilience before deployment.
WonderFence implements dynamic runtime guardrails intended to enforce policy alignment and to block, mitigate, or redirect harmful, non-compliant, or exploitative interactions in live systems.
WonderCheck delivers ongoing automated red-teaming evaluations designed to detect behavioral drift in production, surface emerging risks, and prioritize remediation actions based on detected issues.
The platform emphasizes adaptive, customizable policy alignment across modalities, classifiers and detectors trained on adversarial datasets, and the ability to tune coverage to regulatory requirements and organizational risk tolerance.
Alice also describes continuous telemetry and dataset updates that feed classifiers, guardrails, and red-teaming workflows to maintain coverage against evolving adversarial techniques.
Use cases
Security and safety teams use the platform for pre-launch stress testing of models and conversational agents to identify vulnerabilities such as prompt injection, jailbreaks, and model exploits.
Runtime and engineering teams deploy dynamic guardrails to prevent data leakage, block abusive content, and enforce brand or regulatory policies in chatbots, content pipelines, and automated agents.
Risk, compliance, and product teams run automated red-teaming and drift detection to surface emerging threats, prioritize fixes, and monitor production behavior over time.
Alice lists applicability across industries including child-facing products, financial services, healthcare, insurance, and legal, and across systems such as UGC platforms, GenAI apps and agents, and future robotic integrations.
