AI safety & policy
Information security
Full-time
Salary
$150,000 to $350,000 per year
Routes to impact
Direct high impact on an important cause
Description
Build and own cybersecurity benchmarks for frontier AI models, secure evaluation sandboxes, and shape benchmarking strategy through threat modelling.
- Design, build, and maintain cybersecurity benchmarks, primarily using the Inspect library
- Sandbox and harden evaluation environments so capable AI models cannot escape them
- Prototype new benchmarks and experiments, and keep evaluations running for new model releases
- Use threat modelling to guide what is measured and inform AI cybersecurity evaluation strategy
- Integrate benchmarks with infrastructure and collaborate on accurate evaluation outputs for research
This text was generated by AI. If you notice any inconsistencies, please let us know using this form.
