Transposable elements (TEs) are highly active during vertebrate embryogenesis. The deleterious effects of active TEs have been studied extensively, but despite their selfish origins, once TEs are no longer mobile, their sequences can be coopted by their hosts for a variety of cellular functions. However, whether active TEs can have a developmental function remains an open and provocative question. Zebrafish (Danio rerio) is a powerful system to study embryonic development due to the accessibility of its transparent embryos. Furthermore, the zebrafish genome harbors an abundance and wide diversity of TEs, including many recently active families. Using single-cell RNA-seq data, I identified a small set of TE families with somatic expression during zebrafish embryogenesis, primarily comprising endogenous retroviruses. Among those, I further characterized BHIKHARI (Bik-1), an endogenous lokiretrovirus expressed specifically in mesendoderm cell lineages. I found that Bik-1 is mostly composed of nonautonomous copies which have been transposing very recently but only encode a predicted Gag protein. To knock down Bik-1 expression, I made use of locked nucleic acid (LNAs). Upon LNA injection in zebrafish embryos, Bik-1 mRNA levels were reduced by ~70%, whereas the expression of host genes was not significantly affected. Strikingly, we observed developmental defects in embryos depleted of Bik-1, including smaller heads, lack of eyes and shortened anterior- posterior axis. Rescue experiments indicate that these phenotypes are caused by the depletion of the Gag protein encoded by Bik, likely due to cellular activities mediated by viral-like particles. Another Gag protein from a closely related TE family, BHIKHARI-2 (Bik-2), is also important for the migration of its hosts, the neural crest cells. Taken together, we propose an “addiction” model to explain how organismal development can be become progressively dependent on products encoded by active TEs without fundamentally altering conserved developmental processes. By proving that the mobile and replicative nature of TEs is not incompatible with having a beneficial impact on host fitness, our findings challenge the long-held view that active TEs are merely in conflict with their host. In doing so, it forces a reevaluation of the ‘selfish DNA’ theory proposed four decades ago.