It is fairly concentrated: the five most active assignees together hold 38 of the 49 records in scope, or 77.6% of the field. The top ten reach 93.9%. That leaves a long tail of single-filing entrants — mostly equipment specialists, individual inventors and at least one university — below the tenth position, so most companies entering this space are working around a small set of established filers rather than a crowded open field.

Material analysis and testing (IPC class G01N) is the largest single category, covering 69.4% of the 49 records, followed by image data processing (G06T) at 38.8%. Container and apparatus classes such as A61J, B29C and A61M each cover under a fifth of records. AI-specific classes — G06N for computing models and G06V for image recognition — each sit at 12.2%, the smallest shares tracked, indicating the inspection hardware and methodology layer is far more heavily claimed than the classification software layer.

Based on this dataset, yes relative to the rest of the field: only 12.2% of the 49 records carry a G06N (AI computing) classification and the same share carries G06V (image/video recognition), against 69.4% for the underlying material-testing class. That does not mean the space is empty, but claim density there is markedly lower than around the core inspection hardware and methods, which is worth confirming with a freedom-to-operate search before committing to a specific classifier architecture.