Home / Services / AI & Data Security

Service categories

AI & Data Security

Exchanges with AI, files you receive. Inspect the new entry points into your web services

LLM Firewall inspects input to your generative AI app and its answers before they reach users, and MalScanner analyzes files uploaded by users and files stored in Amazon S3. Prepare for prompt injection, sensitive data being sent to an LLM, malware slipping in and other risks that enter through generative AI apps and file upload features, starting with the service you need.

ContactView services

Services

AI & Data Security services

LLM protection

LLM Firewall

Inspects input to your generative AI app and its answers before they reach users, with a single API. Detects prompt injection, personal and confidential information, web attacks, harmful content and obfuscated input.

File inspection

MalScanner

Automatically analyzes files stored in Amazon S3 and determines whether they are suspected malware. An input-side safeguard for services that accept file uploads from users.

Better together

What changes when you combine them

LLM Firewall + MalScanner

Two entry points: text and files

Input to generative AI apps goes through LLM Firewall, and uploaded files through MalScanner. Both can be called through an API, so you can add inspection at the same point: before passing input to the LLM, and before publishing a file.

LLM Firewall / MalScanner

You decide whether to stop it

Both services return a verdict with reasons. Your app decides whether to block, only warn, or hold and notify the person in charge.

MalScanner → Threat intelligence

Put the traces you find to work

IOCs detected by MalScanner go automatically into your own collection in threat intelligence (Threat Intel Hub), and can be brought into your SIEM or threat intelligence platform over STIX / TAXII.

WAAP + LLM Firewall

Cover the entry points beyond the WAF

Attacks hidden in natural language and instructions concealed by encoding are hard to catch with conventional WAF patterns. In addition to the web entry points WAAP protects, inspect LLM input with LLM Firewall and bring detections into the same console.

Capabilities

Key features by service

Input and answer inspection for generative AI apps

LLM Firewall

  • Prompt injection detection
  • Personal and confidential information detection
  • Web attack pattern detection
  • Harmful content by category
  • Inspection after deobfuscation
  • Reacts to unknown obfuscation
  • Japanese input support
  • Policies, logs and alerts
LLM Firewall details →Malware inspection for files

MalScanner

  • Automatic inspection on S3 upload
  • Format detection by content
  • Per-format analysis
  • Archives inspected inside
  • Verdicts with reasons
  • Mapped to MITRE ATT&CK
  • AI IOC triage and hunting queries
  • API and Threat Intel Hub delivery
MalScanner details →

Getting started

Getting started

  1. Choose the entry points to protect

    Use LLM Firewall for input to generative AI apps, and MalScanner for uploaded files and files placed in S3. Choose to match the entry points you want to protect.

  2. Connect the API or buckets

    With LLM Firewall, call the inspection API just before calling the LLM, using an API key issued in the console. With MalScanner, create a role with CloudFormation and add your S3 buckets, or send files through the API.

  3. Act on the verdict

    Based on the verdict returned, your app decides whether to allow, warn or hold. Results can be reviewed in the console, and LLM Firewall detectors and rules are tuned to your business while you watch the logs.

FAQ

FAQ

Can I contract LLM Firewall and MalScanner separately?

Yes. Each service is contracted individually. If you use both, you manage them with the same account and console.

Does it stop threats automatically when something is detected?

Both services return verdicts. Whether to block or only warn based on an LLM Firewall result, and whether to quarantine or delete a file MalScanner judges malicious, is decided in your app or workflow after you receive the result through the API.

Are the input text and files stored?

LLM Firewall does not store the body of the input text; it records only a hash of the body and the verdict. MalScanner automatically deletes files uploaded through the console or API after 30 days, and only reads and analyzes files in monitored buckets without storing them.

Can I use it outside AWS?

LLM Firewall is an API that does not depend on any particular cloud or LLM, so it can be used even from on-premises systems as long as outbound HTTPS is available. MalScanner's automatic scanning targets Amazon S3 buckets, and you can also send files for scanning through the API or the console.

Service inquiries and demo requests

We will propose the services you need for your environment and challenges.

Contact and demo requestsAll services