The real test here is how it copes with combatants trained using different techniques and equipment. Ai is only as good as its sensors ability to recognise what is happening. The test is too easy when your combatant is the guy who wrote the training manual you based the software on. I'm not saying he's a bad choice but there is a sample size of 1 here which should always ring alarm bells.
No I didn't read the article, its on the daily mail website so not worth it