13 sept 2019
Autonomous ships are a hot topic in the maritime sector; piracy and armed robbery too. Since the beginning of the year, according to the IMB, almost 100 attacks have been committed against ships all around the world. 2018 saw a marked increase in attacks compared to the past few years.
Here is an attempt to describe an autonomous ship hijacking scenario.
In 2025, an autonomous ocean-going container ship is chartered on a busy commercial line between China and Europe. The vessel is a level four autonomous ship according to the IMO classification  . The vessel is underway, and she is now passing through Malacca Strait. All sensors and artificial intelligence (AI) processes are working hard to avoid collision with others vessels, keep water under the keel and maintain the vessel on track .
On the starboard side of the container ship, a small motorboat is steadily heading towards the ship. The vessel’s navigational AI detects the collision course and is beginning evasive maneuveres to avoid the accident: she is coming to her starboard side in accordance with the COLREG regulations. But the small boat constantly heads on the ship, wiping out the container ship’s effort to avoid the collision. This is a pirate craft, and the pirates want to board the container ship.
As there is no more crew onboard, the potential for kidnapping and ransom is obsolete, but there are still motivations for pirates to board a vessel: the cargo, the vessel and the use of those as utilities for smuggling or as a weapon.
The situation of a voluntary collision route has to be understood by the AI otherwise the vessel will probably move erratically, unable to find a proper solution to avoid collision. The AI will have to make a compromise between the ship’s situation and making things worse, potentially leading the container ship to drift, run aground or capsize… It is difficult to predict what the best solution might be to avoid one of these situations.
If the pirates came onboard, what could happen? Of course there is no accommodation as we understand it, but there is still a kind of “bridge” (a place for navigational sensors) and for sure there will be a “AI control room” and an engine room. These three areas are obviously sensitive, and their hijacking or destruction would quickly render the vessel inoperative.
There are actually two solutions: either the AI has been designed in one way or another to handle this type of situation or it has not.
If it is not the case, a priori, it may be assumed that there are no means of action to prevent pirates from boarding an autonomous ship and that, at some point, human supervision may be required to take over the situation. AI has lost.
If it is the case, the issue is much more sensitive since it directly involves ethical considerations regarding the defense of merchant ships against piracy, considerations that are still being discussed at the moment with regard to the protection of manned vessels by human guards. What about an AI that could have at its command even non-lethal weapons to protect the ship? How to be certain of the danger involved? How to be sure about the classification of a person as a pirate? How can the use of force be appropriately adjusted? With which rules of engagement? And once the pirates are finally on board, what action should be taken? AI is lost.
We can see that an event that is plausible  extends far beyond the scope of engineering. The considerations to be made about autonomous ships must not be deprived of a systemic approach to the ship in its environment and in its evolution over time. The technological excitement, very pleasant by the way, should not be allowed to blind us to the imperative need for a thorough reflection on the role of the human being, the seafarer in this case.
Of course humans make mistakes, but they also recover them. On the other side, AI are working on a priori models built to represent the world, but there are captive of them. The limit of autonomy lies more in the supposed representation of the reality of the models used by AIs than in the capacities of the latter. “Fully autonomous ship: The operating system of the ship is able to make decisions and determine actions by itself.”  Those characteristics are considered that are at least necessary to maintain safe and secure navigation.  It has been widely observed over the past 10 years and endemic in some parts of the world.
Thibaut Eude is the founder and CEO of Nostos Systems. He is an former French merchant marine officer and has sailed on various commercial ships. He then developed his operational safety skills and worked as an offshore oil and gas supervisor for more than 10 years. He holds a PhD from the Mines of Paris School of Engineering. His work on the analysis of data from accident investigation reports makes him an expert in the Deepwater Horizon drilling rig accident. He is a French Navy reserve officer as a response and salvage expert and a professor at the French Maritime Academy (ENSM) in the field of human and organisational factors in the conduct of ships (ERM/BRM STCW).