Amazonbot fetches publicly available pages so Amazon services, including Alexa, can reference current information. Its user agent contains Amazonbot, and it follows robots.txt, so you allow or block it the standard way.
It is more of a service and answer crawler than a dedicated training bot, though Amazon notes the data may also feed model training. If you want Amazon's assistants to be able to surface your content, leave it allowed and confirm no WAF rule is blocking it.