Page 205 - Demo

of operational data across propulsion, auxiliary machinery, hull structures, navigation, and safety systems. Historically, only a small proportion of this data was analysed due to technical and human limitations. LR research highlights that traditional empirical approaches typically process around 10% of available vessel data. In contrast, AI-driven models can analyse closer to 90% of it, unlocking patterns and correlations that were previously invisible. In practical terms, this enables predictive maintenance models that track subtle deviations in vibration, temperature, pressure, or energy consumption and flag emerging failure modes well before they become critical. Similar approaches are used to identify abnormal operational behaviours that pose safety or compliance risks, enabling interventions to be planned rather than rushed. LR%u2019s work on data-driven condition-based maintenance demonstrates that these techniques can significantly reduce unexpected failures and narrow the cost gap between older and newer vessels over time. However, LR is clear that AI reliability must be understood in context. These models are only as good as the data they ingest and the governance frameworks that surround them. Bias, data gaps, and poorly defined operating boundaries can all degrade performance. As a result, LR consistently positions AI as a decision-support partner, not a replacement for engineering judgement. The most reliable outcomes occur when AI outputs are combined with domain expertise, physics-based understanding, and structured assurance processes.This is why class involvement is increasingly focused on assuring the AI itself: validating model intent, monitoring performance drift, and ensuring outputs are explainable and auditable. Initiatives such as LR%u2019s Artificial Intelligence Register formalise this approach by providing a structured framework for documenting, governing, and assuring AI systems used in safety-critical applications. Reliability today is high enough to deliver real value, but trust depends on transparency, oversight, and continuous validation.To what extent will real-time ship data affect physical inspections in terms of ensuring compliance with safety and environmental regulations?Real-time ship data is reshaping compliance from a retrospective exercise into a continuous assurance process, with major implications for physical inspections.In particular, as far as environmental regulation is concerned, digital platforms now enable near-real-time monitoring of emissions, fuel consumption, and voyage profiles. Tools within LR%u2019s digital ecosystem allow operators to track compliance with regimes such as FuelEU Maritime on an ongoing basis, rather than assembling evidence after the fact. This improves accuracy, reduces crew workload, and supports earlier corrective action when deviations emerge. From a survey perspective, this wealth of data allows for more targeted, riskbased inspections. Rather than treating all systems equally in every investigation, inspectors can focus on assets exhibiting abnormal behaviour, accelerated degradation, or regulatory risk signals. This approach has already been demonstrated through initiatives such as LR%u2019s collaboration with Latsco, which uses continuous vessel data to inform smarter, evidence-based classification decisions. The result is greater efficiency alongside stronger safety and compliance outcomes.That said, LR is explicit that data does not eliminate the need for physical inspections. Sensors can fail, software can misinterpret conditions, and certain degradation mechanisms %u2014 particularly structural issues %u2014 still require direct human verification. The future model is therefore hybrid: continuous digital oversight combined with strategically deployed physical inspections, informed by evidence rather than driven purely by time intervals. May 2026 201