They do give a clue as to a reason/excuse why not in the article:

Each [AI security] platform surfaces different vulnerabilities, making it difficult to establish a single, reliable source of truth for what is actually secure.

Also, they come up with so many false positives that it's a huge job to check over the reports for something usable.

I asked this at my company wide security training session. The answer I received was that 0 days are hard to detect which is what makes them dangerous. Well duh, you just told me criminals were using currently available open source AI tools to find them. A total non answer was provided.

So I just used the company mandated AI to scan or source code for vulnerabilities and patched the 2 it found.