Scientists recognise three pillars of data considered key to the judgement of whether a substance might cause, or contribute to, cancer, namely laboratory animal experiments, gene disruption assays, and epidemiological studies. These pillars respectively show that the test substance can be linked to cancer in mammalian models, that there's a demonstrable mechanism for cancerous cell formation, and that there are signs of real-life cancers in an exposed population.
In 2015, the International Assessment for Research on Cancer (IARC) examined glyphosate-based herbicides which are used on most GM crops [1]. It found "sufficient" evidence in animal studies, "strong" evidence of cellular and genetic damage of a kind known to induce cancer, and a suggestion of an association with cancer in a large US farm study (the Agricultural Health Study). The latter is on-going, and the authors indicated that the suggested link "should be followed up as more cases occur" over time [2].
Predictably, Big Biotech rushed its damage-limitation machine into action, persuading willing regulators to accept its own, industry-style, science in lieu of the peer-reviewed science assessed by the IARC. The US Environmental Protection Agency (EPA) accordingly pronounced glyphosate "not likely" to cause cancer.