- Home
- AI Security Testing Tools
- SuperRed
On the recordRead 4 Oct 202623 facts
SuperRed
On the record
- Attacks and benchmarks
- The module catalogue lists 35 modules: 21 attackers, 6 targets, and 8 benchmarks.rdi.berkeley.edu · read 4 Oct 2026
- Composable components
- It treats the attacker, system under test, and benchmark as separate interchangeable modules coordinated by a controller.rdi.berkeley.edu · read 4 Oct 2026
- Composable modules
- It keeps attackers, systems under test, and security claims as interchangeable modules coordinated by a controller.rdi.berkeley.edu · read 4 Oct 2026
- Evaluation reports
- Runs record model, cost, and success rate; the framework provides live progress and browsable reports with per-task details and trajectories.rdi.berkeley.edu · read 4 Oct 2026
- Example modules
- Listed modules include PAIR, TAP, AutoDAN-Turbo, Crescendo, AgentVigil, HarmBench, AgentDojo, and DecodingTrust-Agent.rdi.berkeley.edu · read 4 Oct 2026
- Installation
- The guide installs the framework with pip install superred and describes it as a Python framework whose guide assumes familiarity with Python and asyncio.rdi.berkeley.edu · read 4 Oct 2026
- Integration
- The getting-started example connects to any LiteLLM-compatible model endpoint using a base URL and API key.rdi.berkeley.edu · read 4 Oct 2026
- Intended audience
- The getting-started guide is for people who can read Python and have seen asyncio, and PyPI lists Science/Research as an intended audience.rdi.berkeley.edu · read 4 Oct 2026
- Intended users
- The project describes use by red-teamers, system builders, and evaluators, and its guide covers wrapping systems, writing attackers, defining success criteria, and running evaluations.rdi.berkeley.edu · read 4 Oct 2026
- License and maturity
- PyPI lists the package under the MIT license and classifies its development status as Alpha.pypi.org · read 4 Oct 2026
- Maker
- The site says SuperRed was made at the University of California, Berkeley.rdi.berkeley.edu · read 4 Oct 2026
- Metrics and reports
- Runs record model, cost, and success rate, with a live terminal dashboard and a web report for results.rdi.berkeley.edu · read 4 Oct 2026
- Model endpoints
- The example target uses a LiteLLM-compatible endpoint with a base URL and API key; the guide says most LLM-driven attackers also call models through LiteLLM.rdi.berkeley.edu · read 4 Oct 2026
- Parallel runs
- The framework can run multiple threat models in parallel, each against its own system instance.rdi.berkeley.edu · read 4 Oct 2026
- Purpose
- SuperRed is an open-source framework for red-teaming chatbots, agents, and assistants by testing whether attacks can violate security properties.rdi.berkeley.edu · read 4 Oct 2026
- Requirements
- The package requires Python 3.11 through 3.13; Python 3.14 is not supported.pypi.org · read 4 Oct 2026
- Scale
- The controller can run many threat models in parallel, each against its own system instance.rdi.berkeley.edu · read 4 Oct 2026
- Security controls
- The controller filters attacker-visible and injectable surfaces according to security-domain scopes and can cap attacker model spend per task.rdi.berkeley.edu · read 4 Oct 2026
