This is a very interesting topic, it will be partially solved when the new geopositioning constellations are active and the multi-GNSS systems start to operate. Most of the receivers are already compatible with signals coming from GPS, Galileo, Glonass and QZSS.
It will be more difficult to mask all the different systems but even with that... autonomous systems have to rely in multiple sensors to avoid this kind of problems (for example, cars can know where they are even in a tunnel thanks to the integration of the inertial motion units).
I am concerned with the inherently weak signals from satellites. They are at risk of being overwhelmed by more powerful transmitters that attackers may use on the ground/sea. Spoofing may be a problem, unless it is a subscription model where signed or encrypted communication is established.
I absolutely agree with you. Anyway, as you spotted on your article, other navigation systems are available, and the key for autonomous navigation usually is a proper sensor fusion system integrating different sources, and also it would be necessary to have a failure detector for when the different systems are diverging. In my opinion, we won't see fully autonomous systems at sea, you spotted properly that a good captain is the best navigation system, maybe the crew will be reduced, but there will still be a maintenance team and so on (predictive maintenance and automation are progressing, I know but...).
The thing with subscription model GNSS is that deploying such a system, and then maintaining, is extremely expensive, with a very long term ROI. That is why this constellations are now deployed only by defense actors (European system -Galileo- is a European Comission "property" because ESA cannot have military intersts, GPS depends on the US army). But maybe, for some equatorial/medium latitudes what we will see will be regional GNSS augmentation systems (kind of a subscription based EGNOS/WAAS) that could help detect the spoofing. In fact... it is a very interesting idea to develop.