Training
We train organisations in the private and public sectors on the safety of deployed AI systems, the threats those systems meet in service, and the practice of evaluating them.
Programmes are built for the people who carry a system, which in practice means boards and executives, risk, assurance, and procurement functions, public agencies and their supervisors, and the engineers who operate the models day to day.
The material covers how deployed systems fail once they leave the conditions they were developed under, the threats a system meets once it becomes reachable, and the evaluation practice that allows a supplier's claims to be held up against evidence. Each programme is scoped beforehand and taught on the organisation's own systems, in person or remotely, and the methods are the ones we use in our own evaluations.
To commission a programme, write to us under the training route.