Predation can be a very strong selective pressure on prey. Many studies have shown the existence of innate anti-predator responses, mostly in the early developmental stages of juvenile vertebrates. Learning to recognize predators is another possible defensive resource, but such a method involves a high death risk. There is evidence that prenatal learning exists in animals but few studies have explicitly tested for embryonic learning. The aim of this study was to test innate and learned predator recognition in cuttlefish embryos. For this, naïve embryos were exposed to chemical and visual cues emanating from predators, non-predators, and ink. Their response was assessed by measuring their ventilation rate (VR). We first show that VR decreased in response to both visual and chemical predatory cues and ink but not to non-predatory cues. Second, we show that when non-predatory cues (visual or chemical) are paired with predatory cues or ink for several days, embryonic VR significantly decreased. Such a response is likely adaptive, especially in a translucent egg, since it results in reduced movement and hence may lower the risk of detection by visual predators. This freezing-like behavior may also reduce the bioelectric field, thus lessening the predation risk by non-visual foragers. Our results report that cuttlefish embryos had an innate capacity to differentiate between harmless and harmful chemical and visual cues. They were also capable of learning to respond to harmless cues when they were paired with danger (predator or ink) based on conditioning. The combination of these behavioral mechanisms is an example of the early adaptability of cephalopods. Such behavioral plasticity may give the newly hatched cuttlefish a selective advantage when dealing with either known or unfamiliar threats. Nevertheless, more experiments are needed to test the efficiency of the embryos’ response faced with known or new predators.