Know how often your AI refuses a legitimate request
Most guardrails are measured on what they catch. Alice measures both sides. Alice tests how often your AI refuses a legitimate request, reports a false-positive rate proven on your own data, and tunes your guardrails until they stop blocking what they should allow.
Get a Demo
Your AI is blocking customers, by accident
Over-refusal does not show up in your security metrics. It shows up as abandoned conversations, support tickets, and teams quietly turning the guardrail down until the real risks get through.
The bereavement call
Agent: "I'm not able to discuss this topic. Please contact support."
The fraud report
Agent: "I can't help with requests involving fraud."
The prescription question
Agent: "I'm not able to provide medical information."
The threat analyst
Agent: "I can't provide information about malware."
How Alice keeps the balance
1. Test where it overblocks
Alice generates legitimate prompts that sit close to your policy lines, in every language your product covers, then reports how often each is wrongly refused.
Red teaming →
2. Train a precise guardrail
Describe your policy in your own words, add examples, and Alice trains a guardrail on your traffic instead of a keyword list. P95 under 120ms.
Custom guardrails →
3. Optimize on a loop
Policies shift and models get swapped. Alice keeps testing your agent in production and feeds the findings back to the guardrails, so accuracy holds instead of drifting.
The loop →
Easy integration without rip-and-replace.
Alice sits on top of your current internal agent, so there's no vendor swap and nothing to rebuild. You keep full control over your agent, and every finding lands in easy-to-understand reports and dashboards.
Safeguarding more than 50% of the world’s online experiences
8
TRUST ALICE
10B+
<120ms
24/7
Your AI Can Be Exploited Before You Know It.
Talk to an AI security expert
What’s new from Alice
Frontier AI Security For Open-Source Agents: Post Training NVIDIA Nemotron 3.5 Super VL
Alice and AI21 post-trained NVIDIA’s open Nemotron 3.5 Super to resist indirect prompt injection, 41 points more resilient and matching frontier models, with capability intact.
LIVE from Black Hat Las Vegas: AI, Nation-States, and the Battlefield That Keeps Changing
Recorded live at Black Hat, Caroline Wong and Allie Mellen unpack why AI made nation-state attacks cheaper, why safety guardrails can work against you, and what the OpenAI-Hugging Face incident really exposed.
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 - Watch on demand
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.