It says "a geopolitical entity representing a country or dependent territory".
Regardless of that unacceptable definition
There is a particular challenge where ISO has created codes for the same entity as both a Country and a Territory. While we have SameAs links I don't think that's semantically good enough. Otherwise you end up with problems like https://www.linkedin.com/feed/update/urn:li:activity:7365839460372004864/
I think we need to be opinionated as to what really is a country. For example French Guinea is not a country in its own right (isIndependent is false). While we can link both codes to the Territory I don't think we should also classify it as a Country. I know ISO gives it a country code for use in various contexts e.g. sports, domain names but that doesn't mean it is semantically a country for general purposes.

The property is declared on the individual but it is empty.

The instances we have currently are of type IndividualLanguage and LivingLanguage but not of NaturalLanguage, which seems wrong, and may hamper implementation.

This is an issue with the reference data we provide not with the specification document per se.

Rather a site issue than a specification one:
Links to the rdf versions of the ontologies are not usable with "owl:imports", nor may the ontologies be downloaded locally for reuse, because instead of the rdf what is downloaded from the links is the html file.

Change made by ISO 2022-03-03

Changes made by ISO 2022/03/03:
Addition of category region in eng, fra, ara, kur;
Addition of region IQ-KR;
Assign parent subdivision to IQ-AR, IQ-DA, IQ-SU;

with the name-based URI as an alternative using sameAs.
The name-based URIs are unstable (we have had several countries changing name but not ISO code) and the algorithm (e.g. for replacing spaces and punctuation) makes them unpredictable. Most external datasets use the code not the name, requiring an expensive lookup to create a link using the name-based URI.