Discuss how layered controls (AI de-duplication, e-MB, DBT and social audit) strengthen accountability in rural employment programmes. What are their limitations?

VB-G RAM G replaced MGNREGA in all rural areas on 1 July 2026. Under it, logging attendance on the NMMS app does not trigger wages on its own. Each payment must pass several checks: AI de-duplication, e-MB measurement and DBT, with social audit after payment [1]. Each layer plugs a different leak, but genuine workers can still be excluded.

Face-auth attendance → AI de-duplication → e-MB → DBT/FTO → Social audit
   (who came?)          (repeat photo?)  (work done?) (who paid?) (did it happen?)

Fig: Layered verification chain under VB-G RAM G

How the layers strengthen accountability

• AI de-duplication flags repeated or suspicious photos. The wage list and FTO are then withheld automatically until checks are done [1]. This targets ghost workers: about 40 crore of roughly 46 crore person-days were recorded through face authentication [1].
• e-MB links wages to work physically measured by engineers, not just to attendance [1]. This makes the system accountable for output, not only presence.
• DBT sends wages straight to bank accounts after verification. This cuts out cash-handling middlemen and leaves a traceable trail [1].
• Social audit, random district spot inspections and public disclosure add community oversight [1]. States must also prosecute anyone who tampers with the app [1].

Limitations

• Wrong flags shift the cost to workers. A withheld FTO delays honest workers' wages. A Standing Committee found in 2022 that delay compensation was not paid in most places [2].
• Digital exclusion. In 2024 the Committee reported that workers who could not mark NMMS attendance went unpaid. It also found connectivity and smartphone gaps, and that 14% of job-card holders were not Aadhaar-enabled [3].
• Payment-rail failures. Inactive Aadhaar links and frozen accounts have caused failed transfers [2].
• Thin exception route. Only the mate can file an exception. There is no stated deadline for a decision and no way for the worker to appeal [1].
• The checks verify faces, not quantities. An inflated e-MB entry passes even when every face is genuine.
• Opacity and uneven uptake. No false-flag data is published, and only nine states have 100% face authentication [1].

Conclusion. The layers work together: technology checks at the front and community checks at the back. Their strength depends on fair handling of errors. Four steps would help:

• publish false-flag rates
• clear flagged wages within a set time and pay delay compensation automatically
• add a routine Gram Panchayat-level backup for attendance, with a grievance route for workers
• make e-MB checks a focus of social audits

These steps would keep fraud control in line with the Article 41 goal of the right to work and with SDG 8.