Infrastructure for biodiversity that science and policy can act on: Sensors, AI, and the Nature Restoration Law

European biodiversity policy — from the Habitats Directive to the Nature Restoration Law — needs evidence at a scale and quality that fragmented, siloed observation networks cannot deliver alone. LifeWatch-ERIC works as a catalyst and connector: a federated e-Science infrastructure that composes capability where it lives, across eight member countries, three Common Facilities, and a network of partners spanning DiSSCo, eLTER, EMBRC, the ENVRI cluster, EOSC, and the major aggregators (GBIF, OBIS, EMODnet). The connective tissue is FAIR by design — FAIR Digital Objects, the I-ADOPT semantic layer, EcoPortal mappings, and provenance treated as first-class metadata — so that biodiversity observations become evidence that policy can actually cite. This talk, contributed through the Horizon Europe FAIR2Adapt project (which LifeWatch-ERIC coordinates), argues that AI matters most where the human work around biodiversity data is messiest — metadata harmonisation, semantic alignment, workflow composition, and provenance — and that the next frontier is moving inference to the edge so sensing networks stop behaving as passive pipes and start behaving as instruments that know what they observe. Both arguments are grounded in RECUP-DAS, a landscape-scale recovery experiment in the Iberian Pyrite Belt (Huelva, Spain), funded by ERDF Andalusia, where centuries of acid mine drainage are being reversed using DAS technology from the University of Huelva and monitored in real time through a digital twin with LoRa/NB-IoT sensors, edge AI for species ID, eDNA, and FAIR APIs — each event a citable FAIR Digital Object aligned with the Nature Restoration Law. The talk closes on the US/EU collaboration question — pick the protocol, not the platform — and opens three discussion threads for the workshop: hype versus evidence in biodiversity AI, edge-as-instrument without losing scientific accountability, and what a truly FAIR, transatlantic biodiversity commons should look like.