With the increasing permeation of Artificial Intelligence (AI) in all facets of medicine, the expectations of the technology`s potential escalates. Widespread research applications of Computer Vision (CV) and Machine Learning (ML) to minimally invasive surgery promise to enhance clinical perception, support time and safety-critical decision-making making, as well as other applications for AI in education and certification -- ultimately increasing surgical safety. Before these approaches can be translated into clinical practice there is an undeniable need for widespread validation and evaluation of model robustness across distributions related to patient and data diversity. Moreover, in the face of time-critical decisions impacting patient lives, the algorithms` efficiency, correctness, and handling of uncertainty, pertaining to model limitations and subjectivity of ground truth labels, have to be quantified. Due to its highly standardized approach, and globally large case volume, laparoscopic cholecystectomy – the minimally invasive removal of the gallbladder - has become a benchmark procedure for computational exploration of surgical video data. During laparoscopic cholecystectomy, the Critical View of Safety (CVS) offers a well-defined, clinically validated, and universally recognized surgical safety step to ensure the correct identification of relevant anatomy. Failure to achieve the CVS can result in undesired injury critical structures, such as common bile duct injuries, leading to detrimental consequences for patient outcomes and survival. Despite widespread research endeavors offering promising results for computational CVS assessment, there is no translation into clinical practice. The Society of American Gastrointestinal and Endoscopic Surgeons Critical View of Safety Challenge (SAGES CVS Challenge), currently held at MICCAI 2024, targets algorithmic accuracy, robustness, and certainty in CVS assessment during laparoscopic cholecystectomy. The SAGES CVS Challenge offers an unprecedented infrastructure, characterized by multidisciplinary expertise from surgeons, computer scientists, and industry, meticulous data collection and curation resulting in a large and diverse global dataset, and expert consensus-based ground truth labels generated through a quality controlled annotation framework. The extensive CVS Challenge dataset, comprising contributions from 67 surgeons from 57 countries, and all 6 continents, reflects a vast variety of patient, surgeon, and case characteristics required to ensure the real-world applicability of surgical AI. For more details regarding the 2024 SAGES CVS Challenge, we refer to the published challenge proposal and website (www.cvschallenge.org). Simultaneously to achieving algorithmic widespread accuracy in CVS classification, the increasingly high computational barrier to enable widespread deployment of AI systems for surgical safety has to be overcome to fully leverage AI`s potential for intraoperative decision support and risk mitigation. Thus the 2025 SAGES CVS Lighthouse Challenge (CVS Lighthouse) will revisit the CVS classification task, with two main emphases. In terms of uncertainty calibration and robustness, we will focus on how robust are the algorithms when deployed in different conditions (e.g. sites, countries, etc..), as well as how cognizant is the algorithm when its answers might be wrong, to enable using of these algorithms in a safe manner. We will furthermore explore computational efficiency, realizing different tiers of resource settings --- Besides high-end computational resources prevalent in large academic research centers and in industry, the CVS Lighthouse Challenge targets deployment of the resulting technology on desktop computers available to surgeons worldwide. Enabling globally dispersed surgeons - whose original data contributions underwent rigorous data curation, quality-controlled annotation, and stringent model training and evaluation throughout the SAGES CVS Challenge – to use resulting CVS classification technology with the existing resources at hand means closing the cycle from innovative ideas to transformative developments in surgical practice and patient care. To enhance the possibilities of evaluating CVS classification accuracy, robustness and uncertainty calibration, and further emphasize the impact of computational efficiency the SAGES CVS Challenge dataset will be expanded to 1500 videos (50% increase). By adding to the already extensive dataset, we aim to achieve a more balanced distribution, especially emphasizing underrepresented regions. This will be achieved through collaboration with existing data donors from target regions, who initiated prospective data collection during the data acquisition efforts of the initial CVS Challenge. The SAGES CVS Lighthouse Challenge, targeting computational efficiency and uncertainty calibration in sequence to the systematic exploration of accuracy and robustness of computational CVS assessment, has the potential to translate impactful AI applications into widely available patient benefits. Deploying accurate, reliable, and widely generalizable AI in resource mindful way will democratize access to technology-empowered surgical safety.