Glyphosate is the most used herbicide on Earth. After a half-century of use we know only two biodegradative pathways, each of which appears to degrade glyphosate incidentally. One pathway begins with oxidation of glyphosate catalyzed by glycine oxidase (GO). To date, no naturally-occurring GO enzymes preferentially oxidize glyphosate but nonetheless are sufficiently active to initiate its degradation. However, GO enzymes that preferentially oxidize glyphosate over glycine – i.e. glyphosate oxidases – may have evolved in environments facing prolonged glyphosate exposure. To test this hypothesis, we screened a metagenomic library from glyphosate-exposed agricultural soil and identified a glyphosate oxidase from clone 11AW19 (GO19) that prefers glyphosate over glycine by four orders of magnitude. This is the first GO isolated from a natural source exhibiting a glyphosate preference. Not only have we discovered the first glyphosate oxidase in nature, but we have also demonstrated the utility of functional metagenomics to find a glyphosate oxidase with greater catalytic efficiency and specificity than those engineered using directed evolution.