Recent progress in understanding the pathogenicity of multiple sclerosis (MS) has enabled us to develop new drug entities available in the clinic. However, we still have not succeeded in preventing conversion from relapsing–remitting MS to secondary progressive MS (SP‐MS), and in curing this intractable form of MS. Furthermore, diagnosis is usually retrospective, relying on gradual worsening of neurological signs/symptoms. This is due to the lack of understanding of the pathogenicity driving disease progression in MS and of reliable biomarkers reflecting the disease status. Two relevant components are involved in the brain pathology of SP‐MS: neurodegeneration and inflammation. Neurodegeneration can occur spontaneously in a neuron‐intrinsic manner under chronic inflammation, such as glutamate excitotoxicity, mitochondrial/oxidative injury with iron deposit in the brain and loss of trophic support. Meanwhile, inflammatory and/or immune cells stimulated after repeated relapses, including T cells, B cells, and myeloid cells of peripheral and central nervous system origin, could mediate the processes of neurodegeneration. Among them, a higher frequency of leptomeningeal follicle‐like structures observed in SP‐MS patients suggests that immune cells sheltered behind a blood–brain barrier are still active under smoldering central nervous system inflammation. Accordingly, our recent comparative analysis between MS and its animal model, experimental autoimmune encephalomyelitis, raises a new possibility that ectopic expression of eomesodermin in helper T cells constitutes a previously unappreciated subset of helper T cells with cytotoxic potential against neuronal cells. In the present review article, the mechanisms proposed in the pathogenesis of SP‐MS are summarized, and a new pathogenic mechanism for neurodegeneration mediated by unique cytotoxic helper T cells is proposed.