There are two different ALT's involved here - one on the IMG referencing the MAP AREAs and the ALT on each AREA - and there could be null ALTs on either element (or both elements). This poses a few questions:

1. Should the accessible name of an AREA be a combination of both IMG ALT and AREA ALT? That could get very verbose with the example above when reading each AREA link.
2. If not a combination then should IMG ALT be exposed in some other way, such as mapping IMG ALT to AREA link accessible description?
3. What should happen when IMG has null ALT, but IMG USEMAP references MAP AREAs with non-null ALT? Should it set the image to role=presentational, even though the USEMAP attribute says the image isn't presentational? If AREA is voiced as a combination of IMG ALT and AREA ALT then there will be cases where you don't want any IMG ALT to reduce verbosity.
4. What should happen when IMG has non-null ALT, but AREA has a null ALT?

To get an idea of UA differences currently:

• NVDA 2018 with FF60 voices IMG ALT and AREA ALT as link title
• JAWS 2018 with IE11 voices AREA ALT as link title
• JAWS 2018 with FF60 voices AREA HREF as link title
• VoiceOver with Safari on macOS 10.13 voices AREA ALT as link title
• VoiceOver with Safari on iOS 11.4 voices AREA ALT as link title, but only reads IMG ALT (and not AREA ALT) when the speak screen gesture is used (i.e. no link roles read, and no indication AREA links exist - the only way to find the AREA links seems be setting rotor is set to Links, and using the move to next rotor item gesture )