What if we change our perspective? Instead of using it as a definitive identification tool, we could use it as an assistive tool. The idea is to take the data provided by the AI and compare it against the actual specimen in the field.

This would be a reasonable perspective if the application and it's author recognized it's limitations - as in _very obvious_ disclaimers to the end user that it's only intended to be assistive and not definitive.

Instead it claims to be "Safe & accurate", " Professional", "uses AI to analyze your photos and provide comprehensive safety information in seconds.", " expert level identification ", etc.

These statements provide a false sense of confidence to the lay person. An AI system based on image recognition _cannot_ definitively identify certain mushrooms. There simply is no way around that. Visuals are not a reliable species indicator for a lot of fungi, no matter how much compute you throw at it.

Where I live every fall you can find the destroying angel (Amanita phalloides) in city parks. People have died because they have confused them for other mushrooms. My part of the world doesn't have city level mushroom inspectors (that would be awesome). Maybe I'll test this tools claims that it can intact differentiate this one.

Now someone is going to point out that they do have disclaimers! This is true they do, buried much further down the page well below all their marketing bs about providing "expert level identification" (<- this one particularly bothers me as it is a flat out lie. As mentioned above visuals don't cut it. An actual human expert in the field of mycology isn't going to just look at a mushroom to ID it.). So while I'm glad they do have disclaimers somewhere it doesn't seem like the author thinks they are important enough to feature higher up.

Tldr; Sure, tools like this can be assistive. That doesn't mean their authors should be given a free pass to be deceptive and irresponsible in presenting the tools actual capabilities.