When Rhodian Sea Law was written between 600 and 800 AD, the writers made what was then a reasonable assumption: that every ship would carry a crew. Over the next 1200 years, lawmakers worked on the same assumption, building an immense body of laws, conventions and regulations. Now, with artificial intelligence (AI) in use across the industry and autonomous shipping hull-up on the horizon, we have to question how AI will impact maritime law.
What does AI have to do with maritime law?
On the face of it, AI has nothing to do with maritime law—for now. AI is a long way from writing law, although it may assist with research and assessment. However, maritime law needs to work out how to manage AI systems. Autonomous ships are appearing on the coast and in harbours. They’re ships which, to some degree, can operate independently of human interaction. They’re “driven” by AI.
The International Maritime Organisation (IMO) has defined four levels of autonomy:
- Ship with automated processes and decision support: seafarers are on board to operate and control shipboard systems and functions. Some operations may be automated and at times be unsupervised but with seafarers on board ready to take control.
- Remotely controlled ship with seafarers on board: the ship is controlled and operated from another location. Seafarers are available on board to take control and to operate the shipboard systems and functions.
- Remotely controlled ship without seafarers on board: the ship is controlled and operated from another location. There are no seafarers on board.
- Fully autonomous ship: the operating system of the ship is able to make decisions and determine actions by itself.
Autonomous ships aren’t necessarily unmanned, but even manned autonomous ships present challenges under maritime law.
AI and maritime law
Whether we’re discussing international conventions or commercial law, all maritime law is based on the premise that all ships are manned, and people are in control. Understanding the scope of the impact is the first hurdle, so in 2017 the IMO maritime safety committee (MSC) started a regulatory scoping exercise.
But the IMO is only looking at the impact on IMO instruments. The scope of the impact is much wider than that. The UN Convention for the Law of the Sea (UNCLOS) imposes a duty on flag states to ensure adequate manning of their ships; the major contracts for the carriage of good by sea and agreements on limitation of liability impose a duty on the carrier to properly man the ship. In this brave new world, what will that mean?
AI in IMO and UN Conventions
Regulatory Scoping Exercise
In the regulatory scoping exercise, the IMO Legal Committee aims to identify identify provisions in IMO instruments which, as currently drafted, preclude MASS. Once this is done, the IMO will identify common issues throughout. In 2018, the Comité Maritime International (CMI) considered the international regulatory framework for unmanned ships. They looked at:
- SOLAS (Safety of Life at Sea)
- ISM (Safety Management)
- ISPS (Security)
- Polar Code
- MARPOL (Pollution)
- COLREGs (Prevention of Collisions)
- STCW (Training and Certification)
- FAL (Facilitation of Maritime Trade)
- SAR (Search and Rescue)
- SUA (Suppression of Unlawful Acts)
Their position paper and associated questionnaire responses are interesting reading for anyone curious about the topic.
Building on the CMI’s preliminary work, IMO Legal Committee is analysing nine IMO conventions:
1. SOLAS (Safety of Life at Sea)
2. COLREG (Prevention of Collisions)
3. SAR (Search and Rescue)
4. STCW (Training and Certification)
5. STCW-F (Training and Certification for Fishing Vessels)
6. Load Lines (loading and stability)
7. Tonnage Convention
8. CSC (Safe Containers)
9. STP (Special Trade Passenger Ships).
MASS trials are springing up all over the place. Noting this, in 2019 the IMO’s maritime safety committee (MSC) approved brief risk-based interim guidelines for MASS trials. While every journey starts with a single step, there’s a lot further to go in updating international agreements to make way for MASS.
Master and crew
When addressing maritime law and autonomous ships, the master and crew are the first hurdle. Who are they, and do they have to be on board?
The United Nations Convention on the Law of the Sea (UNCLOS) imposes duties on flag states. These duties include ensuring that all ships are in the charge of a master and officers who possess appropriate qualifications, and the crew is appropriate in qualification and numbers for the type, size, machinery and equipment of the ship.
Even for signatories of IMO and UN conventions, those conventions only have power within the framework of national law. In their Responses to the CMI Questionnaire on Unmanned Ships, the Maritime Law Associations of each country discussed whether the master and crew have to be on board the ship. Australia’s response was typical:
“It is possible that the chief on-shore remote controller could be considered to be an unmanned ship’s “master” for the purposes of Australian law, though there is uncertainty.
… National Law states that “master of a vessel means the person who has command or charge of the vessel, but does not include a pilot.”
The uncertainty is whether “command or charge of the vessel” requires presence on board.
… other remote controllers could not constitute the “crew” of a remote controlled unmanned vessel.
…National Law defines “crew” to mean “individuals employed or engaged in any capacity on board the vessel on the business of the vessel, other than the master of the vessel or a pilot”. The crew must be ‘on board.'”
The first IMO instrument aspiring deck officers come across is the International Regulations for the Prevention of Collisions at Sea (COLREGs). We don’t have to go far in the COLREGs before we’re faced with a problem when it comes to autonomous ships. Rule 2a states:
“Nothing in these Rules shall exonerate any vessel, or the owner, master or crew thereof, from the consequences of any neglect to comply with these Rules or of the neglect of any precaution which may be required by the ordinary practice of seamen, or by the special circumstances of the case.”
Does this mean that the vessel itself can take responsibility? Who’s the master of an autonomous ship? If a maintenance team of non-seafarers are doing maintenance on an otherwise unmanned ship, do they need to take responsibility for any mistakes the AI makes in complying with the COLREGs? Anyone who’s been at sea in any professional capacity knows that “the ordinary practice of seamen” is a moving target.
The Rule 3 definition of “in sight” raises the same question as the Rule 5 description of proper lookout: “vessels shall be deemed to be in sight of one another only when one can be observed visually from the other.” Do a camera and a computer count as “visual” or “sight”?
Other instruments face similar challenges. SOLAS requires all ships to be “sufficiently and efficiently manned,” and to maintain a continuous watch. STCW allows the officer of the watch to be the sole lookout during the hours of daylight under certain conditions. When talking about passenger vessels, SOLAS Chapter III, Regulation 10 requires that “there shall be sufficient crew members, who may be deck officers or certified persons on board for operating the survival craft and launching arrangements.”
SOLAS requires masters to report dangers to navigation; both UNCLOS and SOLAS require the master to render assistance to those in distress at sea.
Remote-controlled vessels could one day be controlled from anywhere in the world. At the end of their watch, an operator in Australia could hand over a Panamanian-flagged ship to an operator in India. Is one of those operators the master within the meaning of the rules? Which national laws apply?
Article 6 of the International Salvage Convention raises a related question: if the master has the authority to conclude contracts for salvage operations on behalf of the owner of the vessel, who is the master of an autonomous ship? Can the maintenance team of non-seafarers conclude binding salvage contracts? If STCW only applies to “seafarers serving on board seagoing ships…”then, in the absence of crew serving on board, the current version of STCW wouldn’t apply. So what training will remote operators require? What about certification? The questions just keep coming.
Training and emergency response
People handling ships must be properly trained, of course. That’s why we have the STCW Convention. But even that presents problems. STCW applies to seafarers serving on board seagoing ships. It doesn’t apply to people controlling them remotely. Can we also waive the rules for drills? What would a drill look like on a remotely-controlled ship? Or an unmanned ship? In fact, what will a monitoring station actually do in response to a water ingress alarm on a bulk carrier?
Ship construction, equipment and documentation
Construction rules in IMO Conventions come under SOLAS, MLC and Load Lines. They focus on five main factors:
- Watertight and weathertight integrity
- Fire safety
- Vessel operation and navigation
- Crew safety
- Crew comfort
Regardless of whether a vessel is manned or unmanned, watertight and weathertight integrity doesn’t change. We may need to reassess the rest.
Maintenance personnel will still need safe access to all parts of the ship, but is it really necessary to carry hard copies of all the ship’s paperwork on an unmanned ship? Of the main Conventions, only FAL specifically allows electronic documentation. Even there, do we have to include shore-based crew on the crew list?
Will we still need a bridge and engine control room on board? Navigation equipment? Safety centres on passenger ships? What about firefighters, fire suits and BA sets? Liferafts? We have rules for safe access to vessels. In an emergency, how will emergency teams board the vessel safely? How can an unmanned or remote-controlled ship comply with the requirement to assist vessels in distress? Do we need to rewrite the rules to make electronic distress signals mandatory?
Do unmanned ships still need the same level of ventilation, accommodation and non-toxic materials as manned ships? Do we need to consider the possibility of stowaways on unmanned ships? What about accommodating, feeding and treating survivors from another vessel in distress? None of these hurdles are insurmountable, but they must be addressed. The responses to the CMI questionnaire suggest that, at the moment, the answers would depend on the country.
Inspection, jurisdiction and enforcement
Your Maltese flagged, Belgian owned, Korean built, American managed ship is in international waters, remotely controlled by a Singaporean company using remote workers working shifts from four or five other countries. Which jurisdiction controls the working conditions and training of those workers? If they make a mistake that results in pollution, who is liable? If the pollution was caused by a mechanical or programming failure, who is responsible? Many of these questions are not yet covered by international agreements, leaving shipowners wary.
And it’s not just the shipowners. AI is already helping seafarers, with decision support systems, augmented reality (AR) navigation and more. However, if reliance on an AI’s interpretation of the data contributes to an accident, it’s almost certainly the seafarer who will get the lion’s share of the blame.
Enforcement of the key IMO conventions relies on regular surveys, audits and port state control (PSC) inspections. Surveys and audits are always planned in advance, so the company could send representatives to unmanned ships to assist. PSC is always unannounced, so when a PSC inspector boards an autonomous ship, how will the inspection proceed? A normal PSC inspector requires cooperation from the master, officers and crew. Inspection procedures will need to change, as an autonomous ship might carry only a few maintenance contractors, or no-one at all.
Commercial law is the branch of civil law that covers the conduct of business and trade. In the maritime industry it covers the carriage of goods by sea, contracts, insurance and liability, among others.
Many of the challenges around commercial law and autonomous ships centres on the master. The master is the company’s representative. They’re responsible for the ship’s safety and navigation, and ensuring the ship complies with the multitude of relevant laws and regulations. If the master isn’t on board, who’s in charge? More importantly (to the lawyers), who will take the blame if something goes wrong?
As with international conventions, manning–or the lack thereof–is a key problem. The key agreements governing the carriage of goods by sea (the Hague, Hague-Visby, Hamburg and Rotterdam Rules) all impose a duty on the carrier to ensure a seaworthy ship, and exercise “due diligence” (Hague, Hague-Visby and Hamburg Rules) or take “all measures that could reasonably be required” (Rotterdam Rules) in preventing incidents. In the event of a claim, the courts would have to decide whether AI support systems or autonomous ships work in the carrier’s favour, or against them.
This is a maze. Civil liability depends on the coastal state, flag state, the type of incident, and the nationalities of the people involved. In many countries it’s settled law that a master is vicariously liable for the acts of his servant acting in the course of his employment. If a shipowner is vicariously liable for the acts of the crew, is the shipowner liable for the actions of the autonomous ship? And if so, how will that affect their insurance? Would any of the risk pass to the programmer or the manufacturer?
This will obviously vary depending on the jurisdiction and level of autonomy. Could the shipowner be liable for an accident caused by a software flaw? Would such a flaw affect the seaworthiness of the vessel under law?
Developers of autonomous cars have had extensive public debates about the ethics of various collision scenarios. If a government decides to issue guidance for machine learning models and the resulting compliant model causes a collision, can the government be held liable?
In 2018, the International Union of Marine Insurance noted, “In a global context, the increased automation and the introduction of MASS is expected to reduce the level of risks and casualties within shipping, while at the same introducing risks that have not previously been quantified or insured.”
The CMI found that most jurisdictions adopt a fault-based collision liability. Depending on the jurisdiction, the shipowner and the software manufacture may be liable for a collision caused by an fully autonomous ship. One possible solution would be to introduce strict liability as a fallback position to cover situations where fault isn’t obvious. However, yet again, this would contradict current international conventions on liability and attribution of fault. An in-depth discussion of this topic is beyond the scope of this article, but CORE Advokatfirma and Cefor have published a detailed discussion of civil liability and insurance for MASS.
So, what’s next?
Under UNCLOS, coastal states must not apply “design, construction, manning or equipment [standards] unless they are giving effect to generally accepted international rules or standards.” In the absence of “generally accepted international rules or standards,” for MASS, can coastal states still require a certain level of manning in busy or sensitive areas, such as the Singapore Strait and English Channel?
Sir Bernard Eder suggested creating “some overarching instrument along the lines perhaps of the Polar Code,” and new definitions for generic words such as “master” and “vessel” that would then apply to existing conventions. However, given the speed and agility of international legislative bodies, it’s more likely that states will continue working independently to legislate MASS in their own jurisdictions. By the time a new MASS Convention comes into effect, we’ll have a patchwork of local MASS regulations enshrined in national laws and regional agreements, corresponding fleets of autonomous or remote control ships that can’t legally travel out of their local area, and a web of new contract terms to fill the gaps. Taken individually, none of the issues raised so far are insurmountable hurdles. Taken together, they present an impressive barrier to widespread international MASS.