MIT Technology Review investigation finds over 1,050 migrant deaths since 2015 occurred within range of CBP's AI-enabled border surveillance towers, including 110+ near Anduril's autonomous towers since 2021

MIT Technology Review, working with Times of San Diego, published a data investigation cross-referencing roughly 4,000 locations where migrant remains were found against roughly 600 CBP surveillance tower locations (including Anduril's newer AI-equipped autonomous towers). They found more than 1,050 deaths since 2015 occurred within the towers' advertised detection range, including 110+ deaths since 2021 within range of Anduril's towers specifically. This is a journalistic data analysis of an existing, long-deployed system, not a new product, policy or technology launch.

This is original primary-source analysis, not commentary, and it lands directly on Anduril's border business and CBP's multi-billion-dollar "virtual wall" procurement, which makes it relevant to anyone tracking AI surveillance contracts, border policy oversight, or Anduril's public narrative. It gives advocacy groups, journalists and possibly legislators a concrete, citable dataset to challenge claims that this surveillance saves lives, which could affect funding debates and contract scrutiny. Impact is mostly reputational and political rather than operational: nothing here forces an immediate technical or regulatory change.

The core finding, that deaths occurred "within advertised detection range," is correlational and does not establish that towers failed to detect people or that agents could have reached them in time. Detection range on paper is not the same as actual detection under weather, terrain, vegetation, night conditions or system uptime, and the investigation as summarised does not separate detection failure from response failure. Framing this as proof the "virtual wall" doesn't work overstates what proximity data alone can show.

1. 01Whether MIT Tech Review or Times of San Diego publish the underlying dataset (death locations, tower coordinates, detection-range assumptions) for independent verification, since none is linked in this extraction
2. 02CBP or Anduril's formal response: do they dispute the 1,050 figure, the detection-range methodology, or offer their own detection/response performance data
3. 03Any Congressional, DHS Inspector General, or oversight body follow-up (hearings, audit requests, procurement reviews) citing this investigation within the next 3-6 months