Humans aren't biologically designed to live on plants, and eating foods you aren't adapted to causes harm
Humans are well adapted to plants. We carry extra copies of the gene for starch digestion and have eaten grains, tubers and legumes for hundreds of thousands of years. Most wild plants are inedible, but the ones a vegan diet relies on have long been eaten safely, and the health evidence on well-planned vegan diets is good.
Every animal is adapted to a particular diet. When it eats outside that range it can suffer from poor absorption, toxins or deficiencies. The archaeological record suggests our ancestors ate a lot of meat for a very long time, while many plant foods on our plates are recent, heavily bred and contain compounds we handle only partly. It is reasonable to ask whether a diet built entirely on plants fits the body we evolved.
The premise is right in general: a food an animal can't digest or detoxify can harm it. It fails on the facts for humans and the plants in question.
Humans are adapted to plant food. Populations with starch-heavy diets carry multiple copies of the salivary amylase gene. Starch grains turn up in fossil dental plaque, including Neanderthal plaque. Grinding tools for processing wild grains date back tens of thousands of years. We cannot make our own vitamin C and must get it from food, mostly plants. The 'hypercarnivore' reconstruction cited for the last two million years is one contested reading of the evidence. It is not a consensus among palaeoanthropologists, and it rests heavily on assumptions about what survives in the archaeological record. Meat leaves bones. Plant food mostly leaves nothing.
It's true that most plant species are inedible to us. That tells you nothing about lentils, oats, rice, potatoes or broccoli. Plenty of animals are inedible or toxic too, and nobody concludes from that that meat is dangerous. The toxins in staple plants, such as lectins in raw beans, are well understood and removed by soaking and cooking. People have done that for millennia.
Bioavailability is a fair point in specific cases. Plant iron and zinc are absorbed less well, and B12 needs a reliable source. These are known and manageable. They are not signs of a diet the body can't handle. The test that matters is outcomes. Large cohort studies and the major dietetic associations find that well-planned vegan diets are adequate and associated with lower rates of heart disease and type 2 diabetes.
The point about recent breeding cuts both ways. Modern broiler chickens, dairy cows and pigs are as heavily bred as any vegetable. Today's chickens reach slaughter weight in about six weeks, and nothing like them existed seventy years ago. If newness made a food suspect, it would count against modern animal products too.
Composition: assuming what is true of the parts must be true of the whole. Division: assuming what is true of the whole must be true of each part.
How it applies here: The argument moves from 'most plants on Earth are inedible to humans' to the suggestion that eating plants as a diet is harmful. It treats a fact about plants as a whole category as if it applied to the specific plants people actually eat. Those are a small, long-domesticated subset chosen precisely because they are edible.