How ICARUS is building human rights due diligence into its AI-enabled learning ecosystem
Artificial intelligence is changing education faster than most institutions can update their policies. It can translate learning materials, improve access to knowledge, support educators and create more personalised learning experiences.
But responsible adoption requires a more fundamental question:
How could an AI-enabled education system affect the rights, dignity and opportunities of the people who use it?
At ICARUS, we believe responsible AI cannot be reduced to technical performance, regulatory compliance or a set of ethical principles published on a website. It must be translated into operational decisions: how systems are designed, what data they use, who oversees them, how educators participate and what happens when something goes wrong.
This is why ICARUS is developing a human rights due diligence approach aligned with the UN Guiding Principles on Business and Human Rights and our participation in the UN Global Compact.
From business risk to risk to people
Traditional risk management generally begins with the organisation: financial exposure, legal liability, cybersecurity, reputation or operational disruption.
Human rights due diligence begins somewhere else—with people.
For an AI-enabled education organisation, those people include learners, employees, educators, professor contractors, institutional partners and communities affected by the technologies and infrastructure supporting digital learning.
The objective is not to suggest that harm has occurred. It is to identify how harm could arise, determine which potential impacts may be most serious and establish measures to prevent, mitigate and remedy them.
The ICARUS potential-impact framework
As part of this work, ICARUS has developed a working register of 15 potential impact areas across four connected dimensions.
1. Responsible AI and learner rights
This dimension considers potential AI bias, inaccurate or harmful outputs, inadequate transparency, insufficient human oversight and barriers preventing learners from questioning AI-influenced outcomes.
It also addresses accessibility, digital exclusion and the protection of children and other potentially vulnerable learners.
2. Privacy, security and safeguarding
Educational technology may process personal, behavioural and learning-related information. Responsible governance therefore requires clear purposes, data minimisation, appropriate access controls, secure infrastructure and understandable privacy information.
Safeguarding must extend beyond technical security. Learners need accessible channels through which they can ask questions, raise concerns and receive meaningful support.
3. Decent work and educator voice
ICARUS works with an international network of approximately 170 remote professor contractors. Human rights due diligence must therefore consider contract clarity, payment, workload, scheduling, wellbeing, equality, academic freedom and access to confidential grievance mechanisms.
Educators should not be treated merely as content suppliers. Their professional judgement and lived experience are essential to the responsible development of AI-enabled education.
4. Institutional, cultural and environmental impacts
AI-enabled education also operates through relationships with technology vendors, cloud providers and institutional partners.
Potential impacts may arise from partner misuse, unequal institutional access, cultural or linguistic misrepresentation and the environmental footprint associated with cloud infrastructure, data storage and AI computing.
These impacts may occur outside an organisation’s direct operations, but they should not automatically fall outside its field of responsibility.
Severity before probability
Human rights due diligence differs from conventional corporate risk scoring in an important way.
The first consideration is the severity of a potential impact on people. Likelihood remains relevant, but a potentially severe impact should not be ignored simply because it appears unlikely.
For ICARUS, the initial priority areas include:
- responsible AI, bias and human oversight;
- learner privacy, security and safeguarding;
- accessibility, inclusion and linguistic equity; and
- decent work, voice and remedy for professor contractors.
These priorities represent a working assessment that must continue to be tested through evidence and engagement with affected stakeholders.
Engagement must influence decisions
A human rights framework designed only by management will always be incomplete.
ICARUS therefore intends to deepen engagement with learners, professors, employees and institutional partners. This may include confidential surveys, listening sessions, accessibility testing, multilingual feedback, responsible-AI scenario testing and consultation with credible representative organisations.
The purpose is not simply to demonstrate that engagement occurred. It is to understand people’s experiences and document how their input changes policies, products and operational decisions.
From principles to accountability
Responsible AI becomes meaningful when responsibility is assigned.
ICARUS’s developing approach connects executive oversight with responsibilities across technology, learning, content delivery and partner management. It also calls for:
- an inventory of material AI systems and use cases;
- human rights and responsible-AI impact assessments;
- data and vendor due diligence;
- accessibility and safeguarding reviews;
- risk-based contractual requirements;
- measurable action and evidence registers;
- confidential reporting and non-retaliation protections; and
- periodic review by senior management.
The objective is to move from general commitments to evidence that systems are operating as intended and improving outcomes for people.
Remedy is part of responsible innovation
Even strong preventive controls cannot guarantee that every problem will be avoided.
Responsible organisations must therefore provide accessible ways for people to raise concerns. They must investigate concerns impartially, correct adverse outcomes and learn from what occurred.
In education, remedy may involve correcting information, reconsidering an AI-influenced outcome, restoring access, protecting personal data, addressing a payment issue or changing a process that created exclusion.
A grievance mechanism should not be viewed merely as a compliance channel. It is also an important source of institutional learning.
A continuing journey
ICARUS does not present this framework as a declaration that every risk has been resolved. Human rights due diligence is an ongoing process that must evolve as technologies, partnerships, markets and learner needs change.
Our direction is clear: AI-enabled education should expand human capability without diminishing dignity, agency or equality.
The future of education will not be judged only by how intelligently its systems perform. It will be judged by whether those systems help people learn safely, participate meaningfully and exercise greater control over their futures.
That is the standard responsible educational innovation must meet.
Sources:
- UN Guiding Principles on Business and Human Rights
- Foundation for human-rights due diligence: identify, prevent, mitigate and account for adverse impacts. Principle 24 supports prioritising impacts according to severity. UNESCO Recommendation on the Ethics of Artificial Intelligence
- Supports human dignity, fairness, privacy, transparency, accountability, human oversight, sustainability and the principle of doing no harm. UNESCO Guidance for Generative AI in Education and Research
- Supports human-centred, age-appropriate, equitable and privacy-protecting adoption of generative AI in education. UNESCO Artificial Intelligence in Education
- Provides context on AI’s educational opportunities alongside risks to inclusion, equity and human agency. UNESCO AI Competency Framework for Teachers
- Useful for claims about educators’ agency, professional development, ethical awareness and the continuing importance of teachers. UNICEF Guidance on AI and Children
- Covers child-centred AI, safety, privacy, non-discrimination, transparency, accountability, inclusion and children’s best interests. UN Global Compact Business and Human Rights Navigator
- Practical support for identifying and addressing human-rights risks across operations, supply chains and business relationships. OECD Due Diligence Guidance for Responsible Business Conduct
- Supports risk-based due diligence covering workers, human rights, consumers, the environment, business relationships and remediation. EU Artificial Intelligence Act – Regulation (EU) 2024/1689
- Authoritative legal source for the EU’s risk-based AI framework, including transparency, data governance, oversight and safeguards. EU General Data Protection Regulation
- Supports the article’s discussion of lawful processing, data minimisation, security, transparency and individual data rights.