AI-Powered Mitigation Suggestions

Intelligent Security Guidance

Optionally connect your own LLM to get context-aware remediation guidance. ThreatMitigator never sends your source code to AI providers—only redacted threat metadata.

Supported Providers

Provider	Models	Setup
OpenAI	GPT-4o, GPT-4, GPT-3.5-turbo	`OPENAI_API_KEY` env var
Anthropic	Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku	`ANTHROPIC_API_KEY` env var
Ollama	LLaMA, Mistral, CodeLlama, any local model	`OLLAMA_HOST` env var (default: localhost)

AI Features

Interactive Queries

Ask follow-up questions about specific threats:

1
threatmitigator query THREAT-ID "How do I fix this?"

AI-Enhanced Reports

Add --ai-enhance to any scan for detailed remediation steps with code examples:

1
threatmitigator scan terraform ./infra --ai-enhance --format pdf

AI-Adjusted Severity

LLM analysis can refine severity scoring based on deployment context, providing more accurate risk assessments tailored to your environment.

Fully Optional

All core detection works without AI. The LLM adds depth, not coverage. Rule-based detection is always free and requires no external services.

Privacy-First Architecture

What Gets Sent to AI

Data Type	Sent to AI?
Resource types	Yes (needed for analysis)
Security patterns	Yes
Resource names	No (redacted to `[REDACTED]`)
IP addresses	No (redacted to `[NETWORK_RANGE]`)
Source code	Never

AI Security Hardening

Rate limiting and request budgeting
Secret management via secrecy crate (zeroized on drop)
Response validation against schemas
Retry logic with exponential backoff
Graceful degradation if AI is unavailable

Configuration

Configure your preferred LLM provider in .threatmitigator.toml:

1
2
3
4
5
6
7
8
[ai]
provider = "anthropic"  # or "openai" or "ollama"
model = "claude-3-5-sonnet-20241022"

# Ollama-specific settings
[ai.ollama]
host = "http://localhost:11434"
timeout = 60

Provider Setup

OpenAI:

1
2
export OPENAI_API_KEY="sk-..."
threatmitigator query T-001 "How do I fix this?"

Anthropic Claude:

1
2
export ANTHROPIC_API_KEY="sk-ant-..."
threatmitigator query T-001 --provider anthropic

Ollama (Local):

1
2
export OLLAMA_HOST="http://localhost:11434"
threatmitigator query T-001 --provider ollama --model llama3

Cost Considerations

AI features use your own API keys, so costs vary by provider:

OpenAI GPT-4o: ~$0.005 per query
Anthropic Claude Sonnet: ~$0.003 per query
Ollama (local): Free, requires local GPU/CPU resources

ThreatMitigator’s AI integration enhances security analysis while keeping you in complete control of your data and costs.