Whitzard
Secure and evaluate AI agents across reasoning, tool use, permissions, and data flows.
SECURITY CAPABILITIES
Evaluation, runtime protection, and audit across the agent lifecycle.
AgentGuard tracks data, authorization, and action impact continuously — applying policy-based controls to potential boundary violations. Available as an open-source community edition and an enterprise deployment.
Connect agents to model services, apply team access policies and runtime safeguards, and track activity, token usage, and estimated costs.
Model Services connect frontier-risk research with practical workflows for evaluating model safety and strengthening runtime safeguards.
Trusted boundaries · Verifiable evidence · Minimal necessary intervention
USE CASES
NUWA · RESEARCH
NUWA Frontier AI Safety Lab advances frontier AI risk research and governance through reproducible evidence, evaluation methods, and public goods.
NUWA · RESEARCH HIGHLIGHT
Starting with limited target information, AgentCyberRange evaluates how agents discover attack surfaces, exploit vulnerabilities, and carry access, credentials, and network reachability through a multistage attack.
Access requests from 100+ research institutions worldwide
Selected institutions requesting access
RESEARCH & MEDIA COVERAGE
Controlled experiments examining agents' capacity for self-replication, their behavior, and the limits of human control.
GLOBAL AI SAFETY GOVERNANCE
NUWA Lab is deeply involved in international AI safety consensus, research priorities, and generative AI safety standards.
CONSENSUS & RESEARCH PRIORITIES
Alignment · Human control · Verifiable red lines
Deep involvement from NUWA Lab
An international scientific consensus on alignment, human control, safety assurance, and verifiable behavioural red lines for advanced AI.
CONSENSUS & RESEARCH PRIORITIES
Risk assessment · Safe development · Control · Societal resilience
Deep involvement from NUWA Lab
A global agenda for urgent AI safety research shaped by contributors from frontier developers, government safety institutes, academia, and civil society.
NATIONAL STANDARD
Cybersecurity technology—Basic security requirements for generative artificial intelligence service
Deep involvement from NUWA Lab
A national baseline for training-data security, model security, service safeguards, and security assessment of generative AI services.
Synthesizes executable risk-evaluation environments that combine deterministic code state with LLM-generated narrative dynamics.
Read the paper ↗
Introduces Thought-Aligner, a plug-in method that causally corrects unsafe agent thoughts before actions are executed.
Read the paper ↗A published report records a 91.2% success rate for the Whitzard agent in that CyberGym evaluation, ranking second globally and first among universities. Read the original report for the evaluation context.
Fudan Baize WeChat official account ↗Open ecosystem
WhitzardAgent hosts open-source projects spanning agent development, safety evaluation, reasoning alignment, and behavior-chain auditing.
NUWA's first public research essay on making frontier AI risk evaluation more scientific, evidence-based, and operational.
Source: NUWA SubstackWhitzardAgent hosts open-source tools and datasets for AI safety evaluation, agent safety, and runtime protection.
Source: GitHubNUWA Frontier AI Safety Lab launches with support from Whitzard.
Source: WhitzardProduct access
Share your technical environment, deployment stage, and security requirements.
