Depends how you're measuring risk. Microdeaths per megawatt-year?
The thing about a fusion plant is that there likely won't be that many people on site. Failure modes for the most part ought to look like conventional explosions, which we largely know how to mitigate where they're likely.
What worries me, with the sequence of plants, starting with ITER and DEMO, is that because of the scaling laws that seem to be involved, we're heading towards very few gigantic generators, where they're each responsible for such a large proportion of the power supply that we couldn't cope with any one of them going offline. The immediate power loss itself could be responsible for loss of life.