The welfare of animals in captivity is of considerable societal concern. A major source of stress, especially for wild animals, is the lack of control over their environment, which includes not being able to avoid contact with human beings. Paradoxically, some studies have shown that interactions with human beings may improve the welfare of wild animals in captivity. Here, we investigated the behavioural (behaviours indicative of cooperation or stress) and physiological (variations in salivary cortisol concentrations) effects of the increasingly used practice of training wild animals as a way to facilitate handling and/or as behavioural enrichment. We evaluated the effects of indoor training sessions with familiar caretakers on nine human-socialised individuals of a wild species, the wolf (Canis lupus), in comparison to nine individuals of its domesticated form, the dog (Canis lupus familiaris). All animals were raised and kept in intraspecific packs under identical conditions—in accordance with the social structure of the species—in order to control for socialisation with human beings and familiarity with training. We also collected saliva samples of trainers to measure GC and testosterone concentrations, to control for the effects of trainers’ stress levels on the responses of the animals. During the training sessions, separated from pack members, the animals stayed voluntarily close to the trainers and mostly adequately performed requested behaviours, indicating concentration to the task. Similarly to dogs, the salivary cortisol level of wolves–used as an index of stress—dropped during these sessions, pointing to a similar stress-reducing effect of the training interaction in both subspecies. The responses to the requested behaviours and the reduction in salivary cortisol level of wolves and dogs varied across trainers, which indicates that the relaxing effect of training has a social component. This points to another factor affecting the welfare of animals during the sessions, beside the rewarding effect of getting food and control over the situation by successfully completing a task. As all responses performed by the animals corresponded to cues already familiar to them, the reported effects were likely due to the above cited factors rather than to a learning process. Our results support previous findings suggesting that training is a potentially powerful tool for improving welfare in some wild social canids by creating structured and positive interactions between these animals and their human caretakers.