The day-to-day work of primary care (PC) was substantially changed by the COVID-19 pandemic. Teaching practices needed to adapt both clinical work and teaching in a way that enabled the teaching process to continue, while maintaining safe and high-quality care. Our study aims to investigate the effect of being a training practice on a number of different outcomes related to the safety culture of PC practices. PRICOV-19 is a multi-country cross-sectional study that researches how PC practices were organized in 38 countries during the pandemic. Data was collected from November 2020 to December 2021. We categorized practices into training and non-training and selected outcomes relating to safety culture: safe practice management, community outreach, professional well-being and adherence to protocols. Mixed-effects regression models were built to analyze the effect of being a training practice for each of the outcomes, while controlling for relevant confounders. Of the participating practices, 2886 (56%) were non-training practices and 2272 (44%) were training practices. Being a training practice was significantly associated with a lower risk for adverse mental health events (OR: 0.83; CI: 0.70–0.99), a higher number of safety measures related to patient flow (Beta: 0.17; CI: 0.07–0.28), a higher number of safety incidents reported (RR: 1.12; CI: 1.06–1.19) and more protected time for meetings (Beta: 0.08; CI: 0.01–0.15). No significant associations were found for outreach initiatives, availability of triage information, use of a phone protocol or infection prevention measures and equipment availability. Training practices were found to have a stronger safety culture than non-training practices. These results have important policy implications, since involving more PC practices in education may be an effective way to improve quality and safety in general practice.