How the Crescent Council is Planning to Redefine AI Ethics and Skilling in India

AI is now essential on India’s campuses. It assists with lab reports, timetables, recommendation engines, and research evaluation. However, without oversight, it poses significant risks to privacy, ethics, and social stability.

The Council for AI Ethics and Skilling (CAES), proposed within the Crescent Ecosystem, will guide AI adoption in Indian universities, engineering colleges, and research institutions. Its mandate is to shape the use of AI, ensure equitable benefits, protect stakeholders, and support innovation. CAES will minimise regulatory, ethical, and reputational risks.

CAES will pioneer a pragmatic, human-centric approach to AI. It will prioritise safety, diverse representation, and cost-effective scalability. Rather than costly, fragile automation, CAES will keep AI as a tool for human decision-making. It will not replace human judgment.

The “Human‑in‑the‑Loop” Advantage

AI as a Tool, Not a Replacement

CAES’s core principle is that AI must use a “human-in-the-loop” (HITL) approach by default. Here, humans remain actively involved in reviewing, approving, and contextualising AI outputs. This is especially important for decisions that affect lives, careers, or academic futures.

In practice, this means:

• AI can recommend, but humans decide.
• AI can draft, summarise, and analyse, but humans sign off.
• AI can surface patterns, but humans interpret and act on them.

This approach addresses the risk of delegating important decisions to opaque models. In education and research, where outcomes impact grades, scholarships, careers, and research integrity, AI should not be the final decision-maker.

Mitigating Risk and Cost

Many institutions underestimate the complexity and fragility of highly autonomous AI projects. Fully integrated systems require extensive connections to ERP platforms, learning management systems, HR databases, and security layers. Such systems often:

• Take years, not months, to deploy.
• Accumulate high consulting, integration, and licensing costs.
• Such systems often take years, not months, to deploy. They accumulate high consulting, integration, and licensing costs. They can fail quietly when models drift, governance lags, or data quality is poor.

CAES will instead implement 4-to-8-week, human-supervised pilot projects.

• Each pilot will begin with a focused use case, such as assisting faculty with feedback, helping students explore research literature, or supporting administrators with documentation.
• Pilots will keep integration with core production systems to a minimum. Data flows will be carefully controlled.
• A human reviewer will be required for all high-impact outputs.

If a pilot fails, financial and reputational risks remain limited. If it succeeds, the institution gains practical evidence before scaling. This staged approach reduces risk and helps keep institutions competitive.

Accountability Where It Matters Most

Accountability cannot be delegated to algorithms. For high-impact domains, we set clear boundaries.

• Grading and assessment: AI may assist with rubric alignment, draft comments, or highlight anomalies, but final grades must be assigned and justified by human faculty.
• Hiring and promotion: AI can help filter CVs, summarise portfolios, or highlight patterns, but hiring committees and academic boards retain full responsibility for decisions.
• Research evaluation: AI tools can map citations, cluster topics, and identify likely duplicates or plagiarism, but only human experts can assess originality, impact, and integrity.

This design keeps humans in control. Responsibility for fairness and quality stays with individuals, not platforms.

Why Diverse, Non‑Technical Voices Matter

Ethical AI is often seen as a technical or regulatory issue. While technical and legal expertise are necessary, they are not sufficient on their own.

AI systems operate within societies and shape access to opportunities, language priorities, historical preservation, and social challenges. Metrics or compliance checklists alone cannot capture these impacts.

For this reason, CAES intentionally brings together:

• Technologists and data scientists.
• Cybersecurity and infrastructure experts.
• Legal professionals and policy specialists.
• Sociologists, ethicists, educators, and community representatives.

Each group identifies unique risks, trade-offs, and blind spots. Diversity is seen as an asset, not a barrier.

Spotting Systemic Bias Before It Scales

AI can marginalise vulnerable groups by direct discrimination. It can also create subtle, structural disadvantages.

• Training only on English‑language or urban datasets, leaving rural or regional realities underrepresented.
• Prioritising metrics like “efficiency” or “completion rates” without considering socioeconomic context.
• Using opaque risk scores in admissions or scholarships that reproduce historical bias.

Non‑technical voices are crucial in asking the questions that code alone cannot answer:

• Whose data is missing?
• Who might be harmed if this model becomes standard practice?
• How will this system be perceived by the communities it affects?

By formalising roles for social scientists, ethicists, and community advocates, CAES aims to identify systemic issues early in the design and governance process, rather than responding to reputational damage after deployment.

Agility through Technological Neutrality

Pivot for Accessibility

Technological neutrality means not favouring any specific technology or vendor. To lower costs, enhance security, and improve manageability, CAES will defer neutrality to later phases. Some institutions may be excluded at first. Once a system is established, CAES will adopt broader neutrality rather than mandate a single stack. This means:

• Security and governance baselines that apply regardless of the underlying platform.
• Architectural patterns that can be implemented on multiple infrastructures.
• Interfaces and processes that are agnostic to vendor or OS (operating system), as long as they meet the ethical and security bar.

This approach preserves the benefits of common reference architectures while avoiding rigid requirements that could hinder adoption.

Cost Control without Compromising Safety

Technological neutrality is both a philosophical and financial consideration. Many engineering colleges and universities:

• Run mixed environments (Windows labs, Linux servers, cloud‑hosted services).
• Face tight budgets for infrastructure refresh cycles.
• Need to show results quickly to justify further funding.

With CAES, institutions will be able to reuse upgraded infrastructure, security controls, governance, and AI practices. This will help them:

• Avoid unnecessary rip‑and‑replace projects.
• Focus budgets where they matter most: training people, building governance capacity, and developing open, reusable AI content and tools.

In summary, the Council’s neutrality will enable agility and accountability. Institutions could use their current infrastructure as long as they meet ethical and security standards.

Fueling the Vision: A Call for Grants and Partnerships

Scaling the Ecosystem

CAES will serve as a policy body and catalyst for India’s AI talent and research capacity. Its roadmap includes:

• Integrating AI into core science curricula in a way that is hands‑on, ethically grounded, and aligned with national regulatory bodies.
• Hosting student hackathons and innovation challenges that encourage learners to build socially useful AI solutions under clear ethical and security norms.
• CAES will promote faculty development so educators become confident users and informed critics of AI tools. This approach encourages active engagement rather than passive adoption.

The goal is to spread AI literacy and ethical awareness across all disciplines. This includes civil engineering, biotechnology, humanities, and management. It extends beyond just computer science.

Open for Collaboration

To achieve this vision at scale, CAES will seek grants, partnerships, and opportunities for co-creation with:

• Government ministries and regulatory bodies that care about safe, future‑ready education.
• Corporate innovators who focus on ethical AI can support open, reproducible research. This approach avoids closed black-box deployments.
• Philanthropic organisations focused on digital inclusion, capacity building, and responsible technology.

The goal is to fund:

• Shared labs and reference implementations that multiple institutions can adopt.
• Open educational resources, toolkits, and policy templates.
• Long‑term collaborations should combine academic rigour, industry relevance, and public‑interest accountability.

Our vision is a federated ecosystem where knowledge, tools, and governance models are shared and adapted. We aim for continuous improvement, not isolated efforts by individual colleges.

Conclusion

AI’s role in Indian education and research is evolving. Without oversight, it may lead to surveillance, exclusion, and rigid automation. The Council for AI Ethics and Skilling offers an alternative: AI that is powerful but bounded, innovative yet accountable, and always human-centric.

By promoting human-in-the-loop design, supporting diverse perspectives, embracing technological neutrality, and pursuing collaborative funding and partnerships, CAES will build a foundation for safe, ethical, and inclusive AI.

AI will transform Indian education and research. The question is who will lead and shape that transformation. We invite institutions, policymakers, innovators, and advocates to join CAES in setting a national standard for ethical, inclusive, and human-centred AI. Get involved. Your leadership is essential for building a responsible AI future for India.