AI Specific Attacks and Safeguards

Award outcome

Description

The Term of the Contract is 6 months, with a provision available to extend for an optional 3 months according to AISI's requirements and approval. r the maximum 9-month duration. In delivering this work, a supplier may produce for AISI: Reports and advisory discussions outlining new attacks against frontier AI systems and mitigations that defang them. Code providing novel attacks and corresponding protections, and computational experiments evaluating them. Datasets: Datasets consisting of misuse tasks that frontier AI systems should refuse to execute, for example requests to generate hate speech, write malicious code, create media that incites violence, or synthesise illegal imagery. The Supplier will deliver this work by providing an R&D team that works collaboratively with an AISI researcher whom they meet with once a week. The Supplier's delivery approach should scope precise requirements with AISI's Safeguards Analysis researchers at the project start and subsequently providing a steady stream of relevant output

Links & Documents

About DSIT

Published contracts in the last 12 months

Similar contracts