RNA interference (RNAi)-related pathways target viruses and transposable element (TE) transcripts in plants, fungi, and ecdysozoans (nematodes and arthropods), giving protection against infection and transmission. In each case, this produces abundant TE and virus-derived 20-30nt small RNAs, which provide a characteristic signature of RNAi-mediated defence. The broad phylogenetic distribution of the Argonaute and Dicer-family genes that mediate these pathways suggests that defensive RNAi is ancient and probably shared by most animal (metazoan) phyla. Indeed, while vertebrates had been thought an exception, it has recently been argued that mammals also possess an antiviral RNAi pathway, although its immunological relevance is currently uncertain and the viral small RNAs are not detectably under natural conditions. Here we use a metagenomic approach to test for the presence of virus-derived small RNAs in five divergent animal phyla (Porifera, Cnidaria, Echinodermata, Mollusca, and Annelida), and in a brown alga—which represents an independent origin of multicellularity from plants, fungi, and animals. We use metagenomic RNA sequencing to identify around 80 virus-like contigs in these lineages, and small RNA sequencing to identify small RNAs derived from those viruses. Contrary to our expectations, we were unable to identify canonical (i.e. Drosophila-, nematode- or plant-like) viral small RNAs in any of these organisms, despite the widespread presence of abundant micro-RNAs, and transposon-derived somatic Piwi-interacting piRNAs in the animals. Instead, we identified a distinctive group of virus-derived small RNAs in the mollusc, which have a piRNA-like length distribution but lack key signatures of piRNA biogenesis, and a group of 21U virus-derived small RNAs in the brown alga. We also identified primary piRNAs derived from putatively endogenous copies of DNA viruses in the cnidarian and the echinoderm, and an endogenous RNA virus in the mollusc. The absence of canonical virus-derived small RNAs from our samples may suggest that the majority of animal phyla lack an antiviral RNAi response. Alternatively, these phyla could possess an antiviral RNAi response resembling that reported for vertebrates, which is not detectable through simple metagenomic sequencing of wild-type individuals. In either case, our findings suggest that the current antiviral RNAi responses of arthropods and nematodes are highly diverged from the ancestral metazoan state, and that antiviral RNAi may even have evolved independently on multiple occasions.