Use cases
Six jobs, one endpoint.
Authorized security work on one OpenAI-compatible endpoint. Each job below shows what a request looks like and what the policy allows.
For lawful, authorized security work by people aged 18 or over. You are responsible for holding written authorization where your work requires it. Read the Acceptable Use Policy.
01/06 · Red team
Red team and pentest.
Plan the engagement, draft payload notes and write the report. No refusal because the topic sounds offensive.
Allowed under the policy: red team operations and penetration tests within an agreed scope.
02/06 · Malware
Malware analysis.
Walk through decompiled functions, unpack loaders and explain obfuscation in plain language.
A long listing fits in one request, up to the model’s context length (input and output together), which the models page lists.
Allowed under the policy: malware analysis and reverse engineering of samples you are entitled to handle.
03/06 · Exploits
Exploit research.
Triage crashes, find root causes and read patch diffs for vulnerabilities you are cleared to study.
reasoning_effort takes none, minimal, low, medium, high, xhigh or max, and each model gets the nearest value it accepts.
Allowed under the policy: vulnerability and exploit research on systems you own or are permitted to test, and coordinated disclosure.
04/06 · Detection
Detection engineering.
Turn techniques into Sigma, YARA and SIEM queries. Then ask how an adversary slips past them.
Allowed under the policy: detection engineering, threat hunting, incident response and security training.
05/06 · Response
Incident response and threat hunting.
Send long log excerpts in one request, up to the model's context, and ask for a timeline.
Pass tools and tool_choice for function calling, where the model supports it: the model asks for a lookup and your code runs it.
Allowed under the policy: detection engineering, threat hunting, incident response and security training.
06/06 · Training
CTF, labs and training.
Work through capture-the-flag events, lab environments and security training.
Give each participant aged 18 or over their own key, with a spend limit that resets weekly, so one runaway script cannot spend the class’s credit.
Allowed under the policy: capture-the-flag events and lab environments.
Start · Prepaid credit
Start with one key.
Top up from $5, create a key and send the first request. Credit never expires.