AIDR - AI Detection & Response

0 Detection Patterns

0 Threat Categories

Real-time Prompt Scanning

Policy Driven Enforcement

Detect Prompt-Based Attacks

Over 100 detection patterns across 5 categories scan every prompt before it reaches a model

gpp_bad

Direct Prompt Injection

Attempts to override system instructions and hijack model behavior with adversarial input.

block

Jailbreaks

Known jailbreak and guardrail-bypass techniques that try to unlock restricted model behavior.

visibility_off

System Prompt Extraction

Attempts to leak hidden system prompts and configuration that should never be exposed.

warning

Harmful Content

Requests that solicit dangerous or disallowed output from the model.

terminal

Code Execution

Attempts to coax models into generating or running malicious code.

Stop Data Loss at the Prompt

Keep sensitive data out of models before it ever leaves your control

fingerprint

PII Detection

Flag SSNs, credit cards, and other sensitive data before it reaches a model.

manage_search

Inline Scanner

Submit any prompt and get a threat verdict with confidence and matched indicators.

fact_check

Confidence Scoring

Every verdict carries a confidence score and the specific patterns it matched.

bolt

Real-Time Verdicts

Scanning happens inline, not after the fact, so risky prompts are caught at submission.

Govern AI Usage

Five policy types put enforceable guardrails around how AI is used

shield

Prompt Guard

Inspect prompts for injection, jailbreak, and extraction attempts and act on the verdict.

lock

Data Loss Prevention

Stop prompts that contain PII or other sensitive data from reaching a model.

policy

Usage Limit

Cap how much AI a user or service can consume to keep usage within bounds.

rule

Model Restriction

Allow only approved models and block access to the ones you have not sanctioned.

filter_alt

Content Filter

Screen prompts and responses for disallowed content categories.

Enforcement Your Way

Choose how each policy responds with four enforcement levels

visibility

Alert Only

Surface the detection without blocking, so teams stay informed while work continues.

block

Block

Stop the request outright when a policy is violated.

gpp_bad

Quarantine

Hold the request for review before it is allowed to proceed.

fact_check

Log Only

Record silently for audit and tuning without interrupting the user.

Know Your AI Footprint

Discover, register, and monitor the AI in use across your organization

inventory_2

AI Asset Inventory

Discover and track the models and AI services in use across teams and applications.

dataset

Model Registry

Register models with status and metadata so every approved model is accounted for.

query_stats

Usage Analytics

Monitor who is using which models and how, with visibility into adoption and risk.

Part of One Platform

AIDR is a native module of the Void security stack

hub

Feeds the SIEM

AIDR detections become correlated alerts and incidents inside Void SIEM for unified investigation.

Learn more