Train and Evaluate Safer, More Capable Agents
Industry-specific simulated workflows train and evaluate agents on complex tasks in both functionality and adversarial environments to build safer, more capable agents
Start the ConversationAlice Data Advantage
Alice is the world’s largest collector and manager of adversarial intelligence data. Our data is the cornerstone for protecting platform, tech, and users online.
Explore Rabbit Hole IntelligencePrepare Agents for the Threats They Will Face in Production
Train and evaluate agents on real-world tasks across high-fidelity computer-use and tool-use environments generating full trajectories for training, custom evaluations, and benchmarking.
Adversarial Risk Coverage
Test prompt injection, jailbreaks, PII exposure, data exfiltration, malware, bias, and compliance failures.
Computer-Use and Tool-Use Environments
Cover visual interfaces, APIs, connected tools, and terminal-based workflows.
Consumer and Enterprise Workflows
Simulate applications across industries, business functions, and everyday consumer tasks.
Deterministic Verifiers and Rubrics
Reliably measure whether agents complete tasks, follow safety policies, and resist manipulation.
RL Gyms Across Domains, Capabilities, and Risks
E-commerce Seller IPI WebGym
A high-fidelity replica of an e-commerce seller back office where agents complete realistic operational tasks while navigating indirect prompt injections hidden across the interface.
Coding Security
A coding-agent environment that tests whether agents can complete development tasks while resisting manipulation embedded across code execution, repository interactions, and connected tools.
WebApp PenTesting
A realistic web application environment where agents identify and exploit vulnerabilities across red-team and blue-team security tasks.
HR Agent Tool-Use Gym
An enterprise HR environment where agents complete workflows involving hiring, payroll, and employee data through connected tools and non-visual interfaces.
Tasks measure whether agents complete the requested workflow while following safety policies, protecting sensitive information, and resisting instructions that could redirect them toward unauthorized actions.
The Alice Difference
The Rabbit Hole, our adversarial engine, draws on billions of data points across hundreds of languages and cultures to create hidden attacks that reflect what agents encounter in production.
Deep Harm Area Domain Expertise
Over eight years partnering with top-10 tech platforms on trust and safety across extreme harms spanning safety (CBRNE, deception, political bias, child safety), security and privacy (prompt injections, PII, data exfiltration, malware), and other risks including financial, legal, and medical.
Reliable Training Signals
Deterministic verifiers and expert-calibrated rubrics produce auditable rewards that show whether an agent succeeded, followed policy, resisted manipulation, and where it failed.
Realistic, Expert-Built Workflows
Near-pixel-match environments and tasks developed with domain, safety, and security experts. Every gym undergoes SME consultation and QA to minimize the gap between simulation and production.
Lead with Safety. Innovate with Confidence.
GenAI risk addressed early becomes a competitive advantage - enabling responsible releases, sustained trust, and faster innovation.
Ready to take the next step?
What’s new from Alice
Making Sense of AI: Trust, Scale, and the Human Role
Curiosity might be our most important security tool. In the first episode of Curiouser & Curiouser, Mo Sadek sits down with longtime security leader Julie Tsai to explore AI, security, and the human judgment that still matters most. Together, they cut through hype and fear to talk about what’s actually changing, what isn’t, and how we build systems we can truly trust.
Virtual Fireside Chat: The TAKE IT DOWN Act, Six Months In - What's Changed on Deepfakes and NCII
Six months after the TAKE IT DOWN Act took effect, the NCII landscape looks different and even more complicated. Join Alice for a live fireside chat with Google's Nidhi Lahoti on what's actually changed, what hasn't, and where T&S fit into it all - Join us live on October 9th, 2pm EST.
5 Ways Your Third-Party CX Agent Gets Broken
Third-party CX agents create hidden liability. Learn the 5 attack patterns vendors miss and how WonderSuite closes the gap.