India’s VTU caps AI-generated thesis content at 20 percent

VTU’s 20% cap is an early example of universities moving from hand‑wringing about generative AI to hard numerical rules. It acknowledges that some AI assistance is now baked into research workflows, but tries to draw a bright line between ‘augmented writing’ and outsourcing the core intellectual contribution. In practice, this will force both students and supervisors to develop shared norms and tooling around attribution, detection and acceptable use.

For the race to AGI, this kind of policy doesn’t slow capabilities, but it does shape the human capital pipeline that will work with those systems. If more institutions adopt similar caps, we could see a bifurcation: a group of researchers who still train their own reasoning muscles, and another who lean heavily on models and struggle to build deep intuition. In the medium term that may actually help labs distinguish candidates who can contribute novel ideas to AGI research from those who are primarily good at prompting. Longer term, the bigger challenge is whether such static thresholds stay meaningful as models become integrated into experimental design, simulation and proof‑search, not just prose.