Cyber Evaluations - WP1 - Cyber Range Scenarios

Publication tin regard to award the contract for the provision of CYBER RANGE SCENARIOS - Critical National Infrastructure Additional information: Requirement: Cyber Range Scenarios for Critical National Infrastructure DSIT requires a supplier(s) to design and build realistic computer networks and scenarios that closely resemble the tasks an AI agent would need to complete to conduct an end-end offensive cyber-attack on Critical National Infrastructure. This includes: Specification for a complex Cyber range environment that is that is compatible with our platform, Inspect. Specifications for tasks within that cyber range environment that assess a scaffolded LLM model at cyber relevant tasks. (The construction of an agent is not required). A solver for each task, showing the task is solvable (solutions for all tasks should be provided to AISI). The task details should not be tested on any model with logging. An automated grader for that task that provides details of the success of the model at the challenge. Human baselines and estimates for the expertise level required to solve the task.