News Brief
Machine Unlearning, Synthetic Data Generation, AI Bias Mitigation And More: Government Selects Eight 'Responsible AI' Projects
Generative AI (Representive Image)
The Central government has selected eight 'Responsible AI' projects under the Safe and Trusted AI pillar of the IndiaAI Mission, the Parliament was informed on Friday (21 March).
These projects have been selected to address the need for robust guardrails to ensure the responsible development, deployment, and adoption of AI technologies.
"The projects cover a range of critical themes, including Machine Unlearning, Synthetic Data Generation, AI Bias Mitigation, Ethical AI Frameworks, Privacy-Enhancing Tools, Explainable AI, AI Governance Testing, and Algorithm Auditing Tools," Union Minister of State for Electronics and Information Technology Jitin Prasada said in a written reply to a question in the Rajya Sabha.
The selected projects are:
Machine Unlearning: IIT Jodhpur – Machine unlearning in generative foundation models.
Synthetic Data Generation: IIT Roorkee – Design and Development of method for generating synthetic data for mitigating bias in datasets; and framework for mitigating bias in machine learning pipeline for
responsible AI.
AI Bias Mitigation: NIT Raipur – Development of Responsible Artificial
Explainable AI Framework: DIAT Pune and Mindgraph Technology Pvt. Ltd. – Enabling explainable and privacy preserving AI for security.
Privacy-Enhancing Tools: IIT Delhi, IIIT Delhi, IIT Dharwad, and the Telecommunication Engineering Center (TEC) – Development of robust privacy-preserving machine learning models.
AI Ethical Certification Framework: IIIT Delhi and TEC – Tools for assessing the fairness of AI models.
AI Algorithm Auditing Tool: Civic Data Labs – ParakhAI, an open-source framework and toolkit for participatory algorithmic auditing.
AI Governance Testing Framework: Amrita Vishwa Vidyapeetham and TEC – Track-LLM, transparency, risk assessment, context and knowledge for Large Language Models.
Additionally, the government has launched a second Expression of Interest (EoI) for new Safe and Trusted AI projects, inviting proposals in areas such as watermarking and labeling, ethical AI frameworks, AI risk assessment and management, stress testing tools, and deepfake detection tools.
