Talk at the ADML meetup about Ensuring Safety for Reinforcement Learning.