6 Contributed equally to this work. 7 Corresponding authors Transposons are ubiquitous mobile elements with the potential to trigger genome instability and mutations linked to diseases 1,2 . Antisense piRNAs guide an adaptive genome immune system that silences established transposons during germline development 3 , but how the germline responds to new genome invaders is not understood. The KoRV retrovirus infects somatic and germline cells and is sweeping through wild koala populations by a combination of horizontal and vertical transfers, providing a unique opportunity to directly analyze the germline response to retroviral invasions of a mammalian genome 4,5 . We analyzed genome organization and long RNA and short RNA transcriptomes in testis, liver, and brain from two wild koalas infected with KoRV, while integrating our results with earlier genomic data. Consistent with data from other mammals 6,7 , koala piRNAs were detected in testis and mapped to both isolated transposon insertions and genic and intergenic piRNA clusters. Established transposon subfamilies produced roughly equal levels of antisense piRNAs, which are the effectors of trans-silencing, and sense piRNAs, which drive ping-pong amplification of these effectors 8,9 . KoRV piRNAs, in striking contrast, were strongly sense biased in both animals analyzed. These two koalas each carried 60 germline KoRV-A insertions, but only 14 of the insertions were shared, and none of the insertions mapped to piRNA clusters. The sense piRNAs thus appear to be produced by direct processing of the transcripts from isolated proviral insertions. A typical gammaretrovirus, KoRV produces spliced Env mRNAs and unspliced transcripts encoding Gag, Pol, and the viral genome. KoRV Env mRNAs were 5-fold more abundant than the unspliced pre-mRNAs, but 92% of piRNAs were derived from the unspliced pre-mRNAs. We show that this biased piRNA production from unspliced retrotransposon transcripts is conserved from flies to mice. Retroviruses must bypass splicing to replicate; thus, we propose that failed splicing produces a "molecular pattern" on transcripts from retroviral invaders that is recognized by an innate genome immune system, which silences transposons in cis by processing their transcripts into piRNAs. This innate immune response defends the germline until antisense piRNA production-from clusters or isolated insertions-is established to provide sequence-specific adaptive immunity and memory of the genome invader.The 23-35 nucleotide (nt) long piRNAs bind to the PIWI-clade of Argonaute proteins and guide trans-silencing through base pairing to target RNAs, which leads to transcript destruction and posttranscriptional silencing, or inhibitory chromatin modifications and transcriptional silencing 6 . Transposon silencing by the piRNA pathway is best understood in Drosophila, where the majority