Human Level released its discussion paper Human Rights Due Diligence in the Age of AI: A Discussion Paper on the Potential Uses and Open Questions for AI in HRDD (September 2026).
Human Level’s Take:
- Artificial Intelligence (AI) is here – and it is here to stay. As AI moves from experimentation into the practice of human rights due diligence (HRDD), the conversation has shifted from whether AI belongs in the HRDD toolbox to where and how it can meaningfully contribute and where it may not.
- This comes at a time when we are experiencing two shifts at once: AI use is accelerating, while expectations for effective HRDD are becoming more firmly embedded in regulation and stakeholder expectations.
- Their intersection raises a fundamental question: can AI strengthen the quality, scale and effectiveness of HRDD, or could it accelerate superficial, compliance-driven and overly technocratic approaches that weaken meaningful engagement, accountability and outcomes for people?
- We want to hear from you on our discussion paper. Where are you using AI in HRDD? Where do you see the opportunities? Where are the risks and pitfalls? What can be done to ensure AI contributes to meaningful HRDD?
- Feel free to respond to this email with your thoughts, weigh in on our Linkedin post, or see the discussion paper for other ways of contributing to the discussion.
Some key takeaways:
- The use of AI for HRDD: As AI becomes more capable and accessible, its potential uses are emerging across the HRDD process. From analysing large volumes of information and identifying patterns to mapping supply chains, reviewing grievances and supporting disclosure, its potential uses are growing. This creates opportunities to tackle some longstanding challenges in HRDD, including fragmented information, limited visibility into complex value chains and the volume of information practitioners need to navigate. At the same time, it raises important questions about what effective due diligence looks like when AI becomes part of the process. Where can AI genuinely strengthen the quality of HRDD, and where might efficiency, automation or reliance on available data obscure impacts that are difficult to quantify, poorly documented or experienced by people who are less visible in company systems? Human Level also raises questions around where human judgement should remain central, how meaningful stakeholder engagement can be preserved, how practitioners can interrogate rather than simply accept AI-generated outputs, and how companies can assess whether AI is actually contributing to better outcomes for affected people, rather than simply producing more analysis, more quickly.
- The potential role for AI across HRDD and the open questions in raises at each stage: Human Level maps possible applications of AI against the expectations of the UNGPs and CSDDD, from policy development and embedding through to risk assessment and prioritisation, stakeholder engagement, action and remediation, tracking and communication. In risk assessment, for example, AI could help bring together internal and external information, identify patterns and emerging risks, flag potential under-reporting and support assessments of severity and likelihood, while raising questions about whether available data captures lived experience, context and heightened vulnerability, and whether AI-supported prioritisation remains focused on risk to people. In stakeholder engagement, AI could help identify potentially overlooked groups, tailor information, support translation and synthesise feedback, but may struggle to identify people who are informal, undocumented or absent from existing datasets, or to account for trust, fear, power dynamics and retaliation. These caveats raise fundamental questions about where AI can inform analysis without determining decisions, how its outputs should be tested and challenged, when human expertise and direct stakeholder engagement need to take over, and who ultimately remains accountable for the decisions and actions that follow.
- Effective HRDD, rather than AI capability, remains the entry point for the discussion: Across its closing reflections, Human Level identifies eight themes that recur regardless of where AI is used in the due diligence process. These include the need to recognise that AI can analyse what is visible while some of the most severe impacts may be precisely those missing from company data; to make meaningful stakeholder engagement more important, rather than less, as reliance on AI grows; and to ensure AI augments human judgement rather than gradually replacing it. This also places new demands on practitioners, who will need the capability not only to use AI but to question its outputs and recognise when further expertise, information or engagement is needed. Companies will also need to build greater leverage with AI providers around issues such as data, bias, explainability and confidentiality, while assessing AI’s value not by how much faster or more scalable HRDD becomes, but by whether it leads to more effective due diligence and better outcomes for people. And as AI becomes a tool for conducting HRDD, its use also becomes part of the company’s HRDD picture, bringing the systems, providers and human rights impacts connected to AI itself within the scope of due diligence.
