A working document for assessing human rights risk in AI systems, stage by stage, from first objective to post-deployment monitoring.
The Council of Europe’s HUDERIA sets the expectation: AI systems should be assessed for their impact on human rights, democracy and the rule of law across their lifecycle. It is a methodology, not a worksheet.
As active Observers to the Council of Europe committee guiding HUDERIA implementation, we designed this Workbook as an actionable companion to the framework, a usable instrument teams can complete, developed independently of the Council of Europe. It also produces the documented due-diligence record that obligations like the EU AI Act increasingly require.
Stage-specific questions for planning meetings, design sprints and team check-ins, so human rights thinking enters at each stage rather than as an afterthought.
Structured sections you complete for the record, building a permanent account of your human rights due diligence as you go.
From first intention to long-term accountability, rights considerations enter at each step, not bolted on at the end.
Getting clear on what your ultimate goal is, and building a multidisciplinary team diverse enough to see the risks and the possibilities.
Setting requirements with the communities the tool will affect, and surfacing the trade-offs early.
Asking who is in the data, who is missing, and what history it carries.
Choosing and justifying the model on an ecosystem of values, not performance alone.
Measuring performance across groups, not just in aggregate.
Deploying with sign-off, recourse and ongoing audit in place.
Progressive — work through each stage as you build.
Milestone-based — complete formal sections at project gates.
Comprehensive — run the whole tool at set points, or retroactively on a system already in use. It adapts from a small pilot to a national rollout.