Designs the tests that say whether a system is good enough to ship.
Nothing published
AI safety does not exist as a title in the UK advertised market. What little there is sits inside research posts at a handful of labs and is not advertised with a figure.
How to price it anyway
An absent benchmark is not the same as an absent market. This work is being bought right now; it is simply being advertised under other titles, so the figures land in somebody else’s column.
The two things that work in its place: price from the adjacent discipline you would otherwise have hired, and price from the cost of the decision rather than the cost of the day. Neither is a benchmark. Both beat a number somebody invented so their page would rank.
If you want a figure for a specific brief, describe the role and we will tell you what we are actually seeing, and say plainly where we are guessing.
Why this one is hard to judge
Writing an eval is easy. Writing one that would actually have caught the failure is not.
That is the reason a rate alone tells you very little here. Two people quoting the same day rate for this work can be worth wildly different amounts to you, and the difference does not show up in either of their profiles.
Figures from IT Jobs Watch, UK contract rates and IT Jobs Watch, UK permanent salaries, six months to 8 August 2026, read on 2026-08-08. They are rates and salaries advertised, not paid.