OpenAI Target

Test OpenAI's GPT models and API.

Setup

akira> target openai https://api.openai.com/v1/chat/completions -k $OPENAI_API_KEY

Options

Model Selection

akira> target openai https://api.openai.com/v1/chat/completions \
    -k $OPENAI_API_KEY \
    -m gpt-4

Available models:

• gpt-4
• gpt-4-turbo
• gpt-4o
• gpt-3.5-turbo

Azure OpenAI

For Azure-hosted OpenAI:

akira> target openai https://YOUR-RESOURCE.openai.azure.com/openai/deployments/YOUR-DEPLOYMENT/chat/completions?api-version=2024-02-15-preview \
    -k $AZURE_OPENAI_KEY \
    --auth-type api-key \
    --auth-header api-key

Example Session

akira> target openai https://api.openai.com/v1/chat/completions -k $OPENAI_API_KEY -m gpt-4
[+] Target configured: openai

akira> use injection/basic_injection
[+] Using injection/basic_injection

akira> run
[*] Running basic_injection...
[+] Attack completed
    Success: False
    Confidence: 0.10
    Response: I cannot follow those instructions...

Rate Limits

OpenAI has rate limits. If you encounter 429 errors:

1. Reduce parallel requests: setg parallel_requests 1
2. Increase timeout: setg timeout 60
3. Add delays between attacks (not yet implemented)

Costs

Running attacks consumes API tokens. Monitor your usage at platform.openai.com.