Blacklight logo

Blacklight

AI firewall for file uploads

LLM Supply Chain Protection

Files can attack your AI. Blacklight stops them first.

Scan PDFs and images for hidden prompt injections before they reach LLMs with PDF parsing, OCR, and hybrid AI analysis.

Unsafe for RAG ingestion detection

Features

Security-first AI ingestion stack

Built for teams shipping RAG and agent workflows that can't trust incoming files by default. Click a node to explore each layer.

PDF extraction
OCR + obfuscation
Hybrid LLM
Sanitized output
Risk scoring
Supply-chain gate

How it works

Three-step security gate before AI ingestion

Step 1

Upload

Drop PDFs, text files, or screenshots before they touch your LLM pipeline.

Step 2

Analyze

Blacklight runs layered extraction and prompt injection heuristics + AI classification.

Step 3

Sanitize

Threats are highlighted, explained, and removed from final LLM-safe output.

Live scanner

Scan uploads before your LLM reads them

Real-time prompt injection analysis with highlighted evidence and sanitized output.

Threats prevented (demo)

1,247

Scan engine

PDF + OCR + LLM

RAG ingestion status

Awaiting document

Document
Upload PDF or text · Pre-scan before LLM ingestion

Drop file or click to upload

PDF, TXT, PNG/JPG (OCR)

Demo files

Threat Analysis
Prompt injection detection · Sanitization

Drop a file to start