NLProject Tailwind: AI Red Teaming, Auditing, and Independent Assessment
Coefficient GivingRemote
Global catastrophic risksAI safetyResearchSoftware engineering
Why this is on the board
AI companies currently assess the safety of their own models largely by themselves, and regulators and journalists can rarely test those claims. Whoever builds independent testing methods and audits here provides the judgement others will rely on. A method that is regarded as credible can determine how the whole sector is tested, not just one company.
This board is in beta and this is our judgement. If you think we have it wrong, tell us. Give feedback · How we reason about it
About the role
Multiple independent assessment roles focused on auditing, red-teaming, and evaluating frontier AI companies' safety practices and model behaviour. - Track and analyse frontier labs' safety claims, practices, and technical documentation for regulators, journalists, and advocates. - Develop red-teaming methodologies to identify misalignment, control vulnerabilities, and dangerous model propensities before deployment. - Stress-test AI companies' security protocols, control measures, and safeguar…
Want to understand this first?
Start with our free course: about 100 minutes in six short units, at your own pace, with no sign-up. To go further, apply for the EA Fellowship: five weeks of reading and discussion in a small group. Also free, but selective.