MSC 111. LR × Orca AI. Docker Vision. The IMO Compendium. May 2026 is the month the standards stopped being promises and started being numbers.
Maritime AI is no longer a topic for conferences. It is the operational environment you are working in — whether you are on the bridge, in the engine room, managing a fleet from a Mumbai office, running port operations in Chennai, or advising on maritime insurance in London.
Issue 21 documents where operational evidence and regulatory signals place the standards in May 2026. Not where vendors say they stand. Where class societies, regulators, and verified operational deployments have placed them.
The technology is changing. The professional who understands it changes with it.
MSC 111 is convening in London from 13 to 22 May 2026 with the adoption of the goal-based non-mandatory MASS Code as the defining agenda item, with adoption widely anticipated at the time of publication. Every credible source heading into this session framed it the same way: expected to adopt — the culmination of a process years in the making.
The code does not replace SOLAS or the existing convention structure. It sits alongside it, addressing functions that become safety-critical when human control is remote or removed. The scope covers surveys and certification, approval process, risk assessment, system design, software principles, safe operations, alert management, manning and watchkeeping, navigation, connectivity, remote operations, fire safety, cargo handling, anchoring, machinery, and electrical installations.
The key point for any maritime professional — whether on the bridge, in the engine room, in a shore-based technical role, or in fleet management — is the categorical distinction the code draws.
Advanced automation, integrated alarm management, and voyage decision-support tools are already onboard most modern vessels. None of these automatically place a ship in the MASS category.
West P&I's May 2026 guidance is unambiguous: the only IMO interim guidance currently in force is MSC.1/Circ.1604, and it applies solely to trial operations — not normal commercial trading.
For deck officers, the MASS Code introduces a new watchkeeping and manning framework for vessels where the traditional bridge watch may be remote or reduced. For superintendents and fleet managers, it introduces a new liability and documentation landscape. For port state control officers and surveyors, it introduces a new inspection framework that will take years to fully operationalise.
The correct reading of MSC 111 is not that autonomous ships are arriving on your fleet next year. It is that the industry now has a clearer safety architecture — and the immediate impact for most maritime professionals is interpretive, documentary, and strategic.
Lloyd's Register assessed Orca AI's navigation platform on a feeder containership over 828 nautical miles between Gioia Tauro and Marsaxlokk — through congested traffic, port approaches, and open sea. The evaluation tested object-detection performance alongside radar, AIS, and visual watchkeeping in real operating conditions.
Results as reported in the LR-observed trial: 739 relevant targets across 98 observations, 94% precision, 98.6% recall, 1.4% missed-detection rate, and zero downtime over five days.
What this establishes is a methodology as much as a result. Any AI system seeking class acceptance will increasingly be required to demonstrate measured performance under observation — not cite vendor specifications. For deck officers, the critical question is what happens during the 1.4% — the missed detections — and what the override protocol looks like. For technical superintendents and procurement teams, this trial provides a structured evaluation framework to apply to any AI navigation vendor. For maritime lawyers and P&I correspondents, the liability framework around a 1.4% missed-detection rate in a class-observed trial is a document worth reading carefully.
DNV's new requirement for selective catalytic reduction systems entered into force on 1 May 2026 for ships with keel laid on or after 1 November 2025 or delivery on or after 1 May 2026, under IMO Resolution MEPC.399(83). The former fixed ±5% accuracy threshold for NOx measurement devices has been replaced with a requirement for "sufficient accuracy." The evidentiary burden shifts from a fixed threshold toward demonstrating sufficient monitoring performance over the catalyst's lifetime.
A functional SCR system with poor analyser calibration and an incomplete technical file is no longer sufficient. The evidence trail is what determines whether the vessel demonstrates real NOx compliance — or merely claims it.
For fleet managers and technical superintendents, this is a documentation audit question as much as a technical engineering one.
From 1 May 2026, all EEDI-relevant speed trials must follow ISO 15016:2025 or ITTC Recommended Procedure 7.5-04-01-01.1:2024. ISO 15016:2015 is no longer valid. A technically competent speed trial conducted with an outdated procedure can fail as certification evidence. For shipbuilders, owners, and technical managers, updated methods are not optional — they are the only route to valid results.
Following FAL 49, the IMO expanded its data reference model by adding a Fuel Oil Consumption and CII reporting dataset to the IMO Compendium — over 140 data fields. The same structure supports IMO DCS, EU MRV, CII, and FuelEU Maritime reporting simultaneously.
Noon report discipline, fuel record accuracy, and sensor reliability are now the inputs that determine whether compliance data is usable by AI-assisted reporting tools built on top of this architecture.
The same structural question applies to how performance analytics and reporting are built across the fleet — consistency at source determines reliability at output.
The standardised fields that feed vessel emissions reporting also interface with port call documentation, cargo verification, and customs data flows.
The intelligence in the system is only as reliable as the data flowing into it — and that data originates at every level of the maritime operation.
Every issue, Marine Intelligence Weekly features an Indian company building real technology for the maritime and port sector. In Issue 20, we covered Dtyle.AI and their onboard AI deployment. In Issue 21, we turn to port-side operations — and a Kochi-founded company that has moved well past the pilot stage.
Container trucks arrive in continuous flows. Gate operators manually verify container numbers, ISO codes, seal conditions, and IMDG labels — under time pressure, across shift changes, in dust, rain, and poor visibility. Records in the Terminal Operating System become incomplete. Compliance gaps appear where no individual made a visible mistake — the process itself is the weakness.
This pattern is familiar across the maritime sector. At sea it is deferred maintenance creep — accumulated gaps between what was inspected and what was recorded produce a breakdown that looks sudden but was building for months. At a port terminal, the equivalent is a container moving through a gate without verified IMDG classification, or a damage record that never reaches the cargo insurer. For a fleet operations manager, it is a CII shortfall tracing back to fuel records entered inconsistently across the fleet.
Same structural problem, different operational environment. Docker Vision built its platform to close that gap — not with more personnel, but with AI that sees what the operator needed to see.
When a container truck enters the inspection lane, Docker Vision's dOCR system activates without operator input. Industrial cameras capture the container and vehicle simultaneously. An edge AI processing unit — running locally, not cloud-dependent — performs the full inspection concurrently. The completed inspection becomes a timestamped digital transaction record transmitted directly into the Terminal Operating System.
No operator queuing. No manual data entry. No transcription gap. The truck has cleared the lane.
RFID-dependent systems require pre-installed tags and controlled scan positioning. Conventional OCR reads characters but cannot interpret a partially obscured container marking, adapt to low light and rain, or classify an IMDG label from an oblique angle. Docker Vision's deep learning models understand visual context, not just character patterns.
The following figures are reported by Docker Vision and associated deployment disclosures. Crane OCR deployed in UAE operations consistently delivered over 99% accuracy in live conditions. An Automatic Bag Counting proof-of-concept for PwC achieved 99.99% accuracy, now a commercial deployment processing feeds from nearly 260 cameras at a single site. A gate OCR deployment for a major Indian port group was completed within two days of hardware installation. Across all products, Docker Vision has executed at more than 240 locations.
Gate throughput, IMDG compliance visibility, and damage record completeness. The system adds a verification layer without requiring operators to change workflows.
Container damage records and IMDG verification data feed directly into cargo insurance and liability documentation. Clean gate records reduce disputes and speed claims resolution.
Automated IMDG label detection means hazardous cargo mis-declaration is flagged at the gate rather than discovered at sea. That is a safety intervention, not just efficiency.
The Digital Twin capability provides live operational visualisation — real-time port state that supports congestion prediction and berth planning.
Docker Vision does not oversell the adoption environment. Indian ports vary enormously in infrastructure readiness and digital maturity. Reliability is the central concern — port operations run 24/7 and operators need demonstrated resilience, not presentations. What is shifting is the framing: port operators who previously viewed automation as a cost are increasingly treating AI as operational infrastructure.
This is Article 3 in the Marine Intelligence Weekly India Maritime AI series. Part 1 introduced three Indian companies to watch. Part 2 mapped the full ecosystem — 13 Indian-founded maritime AI companies. This article examines the global companies whose standards every maritime professional will increasingly work within.
In Part 2, we documented thirteen Indian companies building real products — from underwater inspection robotics to predictive maintenance platforms, from shallow-water survey drones to satellite intelligence tools. The picture was one of genuine momentum. But Indian maritime AI is building into a global race already running at speed.
Orca AI uses dedicated low-light cameras and deep learning models trained specifically on maritime traffic data. The May 2025 Series B of $72.5M expanded the platform into naval and defence applications. The LR live trial produced, as reported in the class-observed evaluation: 94% precision, 98.6% recall, and zero downtime across 828 nautical miles.
This is the class-verified benchmark — the question is no longer whether AI navigation works, but what the human override protocol looks like and how missed detections are logged.
The LR trial provides a structured evaluation framework. Apply it to every AI navigation vendor before committing.
The liability framework around a 1.4% missed-detection rate in a class-observed trial is a document worth reading carefully.
SEA.AI integrates thermal cameras, visible light cameras, and radar into a unified AI-processed detection layer. Optical cameras degrade in fog and rain precisely when collision risk peaks — thermal sensor fusion addresses that gap. For deck officers navigating in restricted visibility — monsoon season, North Sea winter, congested straits — this is a fundamentally different capability. For port pilots and VTS operators, the same architecture offers congestion visibility that CCTV and AIS alone cannot provide.
DeepSea's autonomous propulsion optimisation platform adjusts speed, trim, and engine parameters in real time — continuously recalculating the optimal operating point as conditions change. The company has positioned 2026 as the inflection point for propulsion autonomy in cargo shipping. For chief engineers and second engineers managing fuel against CII targets, the shift from advisory to autonomous action is the critical transition — understanding what the system optimises for, where it cannot account for structural loading constraints, and how human override is documented is the practical skill this transition demands.
Cetasol's iHelm delivers real-time fuel reduction recommendations. CetaFuel provides virtual fuel readings without additional sensor hardware — making deployment accessible on vessels where sensor installation is not practical. For HSQE managers and fleet compliance officers, a no-hardware CII monitoring tool changes the deployment conversation: no drydock window, no CAPEX cycle. For engineers managing daily fuel logs, the virtual reading layer adds a cross-check that can flag discrepancies before they compound into a CII shortfall.
Windward fuses AIS data, SAR satellite imagery, vessel behaviour patterns, financial records, and sanctions databases into a real-time maritime domain picture. During heightened regional tensions in the Gulf in early 2026, Windward tracked 686 vessels — identifying behavioural anomalies, AIS manipulation, and sanctions exposure across a fleet actively altering routing and transponder behaviour under pressure.
Windward's intelligence drives the routing advisories and passage plan amendments that arrive from fleet operations during geopolitical events.
It flags your vessels to P&I clubs and port state authorities if AIS behaviour triggers an anomaly.
A primary data source for sanctions exposure assessment and underwriting decisions.
Vessel risk intelligence behind inspection targeting, and real-time supply chain disruption tracking.
Saronic closed an announced $1.75B Series D in April 2026 at a reported $9.25B valuation, following a $392M US Navy production contract. They completed the first hull of their 180-foot Marauder autonomous vessel in under six months. The same route that GPS, AIS, and ECDIS all followed from naval to merchant application applies here. For marine engineers and naval architects, the reliability engineering standards being developed for autonomous vessels are likely to influence the baseline expectation for any unmanned system entering commercial certification.
1. Decarbonisation regulation has no reverse gear — but it has experienced a gear change.
CII, FuelEU, and EU ETS are already in force. The IMO Net-Zero Framework — approved in principle at MEPC 83 in April 2025, delayed at MEPC/ES.2 in October 2025 — is now expected to be formally adopted at MEPC 85 in late 2026, with phased application from 2027. MEPC 84, which concluded on 1 May 2026, advanced the Phase 2 CII review and lifecycle assessment guidelines.
The delay did not reduce compliance pressure. It increased fragmentation.
The log discipline and monitoring architecture you build today is the foundation every future compliance tool will run on.
2. Defence technology is the commercial R&D pipeline. Saronic's announced $9.25B valuation. Orca AI's naval expansion. Windward's domain monitoring role. The boundary between military maritime AI and commercial maritime AI is dissolving. Technologies built for navies have historically reached merchant shipping within five years.
3. Capital is compressing the timelines. Maritime tech startups raised approximately $234M in disclosed funding in the twelve months to mid-2025 — a 73% increase over the prior year (Flagship Founders, Global Maritime Tech Startup Map 2025). The expectation is shifting from pilot programmes toward operational deployment.
The question is no longer whether these technologies will arrive. It is whether the professionals working in the sector will be ready when they do.
From Issue 21, this column broadens its scope. The maritime professional at sea or ashore faces the same fundamental shift — more systems, more data, more accountability for what the systems around them are doing.
The maritime professional of 2026 is no longer judged only by operational competence. They are increasingly judged by their ability to understand, verify, override, and document the systems that assist their decisions.
The LR-Orca AI trial showed that AI navigation can be evaluated objectively — 94% precision is a number any officer, superintendent, or surveyor can work with. The DNV SCR requirement showed that compliance is now an evidentiary question — the shift from ±5% to "sufficient accuracy" places the justification burden directly on the operator and engineer. The IMO Compendium expansion showed that log discipline is becoming compliance infrastructure. The Docker Vision deployment data showed that AI at Indian ports is operational infrastructure, not a pilot programme.
Three things worth carrying from this issue, regardless of where you sit in the sector:
Understanding what a system does and what it cannot do is now a core professional competency — for bridge officers assessing AI navigation aids, for engineers managing automated fuel optimisation, for operations managers using domain intelligence platforms, and for port staff working alongside automated gate inspection.
Log discipline is compliance infrastructure. Whether it is a noon report, a fuel entry, a gate OCR record, or an AIS position update — the reliability of every AI system downstream depends on the quality of the input at the source.
The professionals who will lead this transition are not the ones who know the most about AI. They are the ones who understand their operational domain deeply enough to know where AI helps, where it cannot, and where human judgment remains non-negotiable.
| Topic | What happened | What it means for your role |
|---|---|---|
| Docker Vision | Rs.1 crore USHUS grant, 240+ locations, 99%+ live accuracy (reported by Docker Vision) | Indian port-side AI is operational — relevant for port managers, fleet ops, HSQE, cargo interests |
| MASS Code | MSC 111 convening 13–22 May — adoption widely anticipated | Deck officers: new watchkeeping implications. Shore managers: new liability landscape. All: know the MASS definition |
| AI Navigation | LR-observed trial: Orca AI 94% precision, 98.6% recall, 828nm, zero downtime | Set as your evaluation benchmark — officers, superintendents, procurement teams |
| SCR Compliance | DNV SCR effective 1 May 2026 — MEPC.399(83) | Engine officers: justify "sufficient accuracy." Fleet managers: documentation audit question |
| IMO Net-Zero | Delayed Oct 2025. MEPC 84 advanced guidelines. MEPC 85 late 2026. | EU ETS and FuelEU in force now. Build log discipline and monitoring architecture today |
| Speed Trials | LR: ISO 15016:2025 or ITTC 2024 mandatory from 1 May 2026 | Owners, builders, technical managers: outdated procedure invalidates EEDI evidence |
| Emissions Data | 140+ standardised fields in IMO Compendium post-FAL 49 | Every role touching a noon report, fuel log, or performance report is part of this chain |
| Global AI Companies | Orca AI, DeepSea, Windward, Saronic, Cetasol, SEA.AI expanding in 2026 | Standards these companies set apply across every maritime role — sea, port, and shore |
Maritime AI is in its evidence phase. The professionals who benefit first — at sea and ashore — will be those who understand what the systems around them are doing, can verify their outputs, and know when human judgment must take precedence.
