While machine learning shows immense promise in automated knowledge generation, current techniques like large language models and microtargeted influence operations enable concerning harms like misinformation proliferation. As an exemplar policy response to such harms on online platforms, the EU's Digital Services Act (DSA) warrants comprehensive evaluation of its impacts constraining harmful downstream effects of these opaque practices. Despite harmful applications, we argue machine learning techniques offer immense yet under-exploited potential for unraveling impacts of emerging regulations like the DSA, targeting opaque platform practices enabling misinformation. Following analysis revealing limitations in DSA provisions, we outline a research agenda to strengthen accountability in emerging sociotechnical governance. We call for resolute efforts addressing methodological barriers around appropriate data access, isolating marginal regulatory effects and facilitating generalization across contexts. Given the identified advantages of data-driven approaches to regulatory delivery, we advocate virtuous machine learning research help to quantify policy impacts on online harms.