Artificial intelligence is not something cheap. The deployment of deep learning models requires a lot of hardware if you want to move forward fast. For example, just a few years ago Google spent US$5 billion building an AI-focused supercomputer that today only takes 3 GPU's and costs around $33k.   

Number of cores1K CPUs = 16K cores3GPUs = 18K cores
Training timeweekweek

Since automatic fight detection requires the system to analyse different ways to fight and kinds of a fight, which could take place, it requires almost 4 more hardware (compared to a gun detection). That’s why Athena Security’s Fight Detection system costs 4 times higher compared to our Gun Detection System.