The US Army has awarded Safe Pro Group a $2.5 million contract to integrate its AI threat-detection software into Anduril’s Ghost-X drone. The aim is to help crews spot explosive hazards and drone threats before tanks enter danger zones, though the claimed detection capabilities come from the companies, not an independent test.
Anduril AI Watch analysis
What happened
Business Insider reports that Safe Pro’s Ground Truth software will be integrated with Ghost-X, a small uncrewed aircraft, and feed threat information into military systems including ATAK and Anduril’s Lattice.
Safe Pro says its software can identify and map more than 150 types of explosive threats from aerial imagery. The company says its training datasets draw on Ukraine. The Army contract is for integration; the reported announcement does not establish how accurately the system performs in operational conditions.
Why it matters
This is a concrete military application of AI: analysing drone imagery to locate possible hazards before soldiers and vehicles encounter them. The intended targets include mines, cluster munitions and first-person-view drones that could ambush armoured vehicles.
For Anduril, the project also shows its Ghost-X aircraft being paired with third-party detection software and connected to military networks. That could make the drone more useful as part of a wider reconnaissance system, rather than simply another aircraft in the sky.
Our read
The practical promise is clear: give crews more information before they move into danger. But “can identify” is a company claim, not a field result. The meaningful test will be whether the system can reliably distinguish threats in difficult conditions, and whether the information reaches the people who need it in time. Battlefield technology has little patience for a polished demo.
What to watch
- Whether the integration moves from contract work into operational use.
- What evidence emerges about detection accuracy across different threat types and terrain.
- How the Army and Anduril describe the system’s role in protecting crews and supporting reconnaissance.
Discussion spark: Should battlefield AI systems be judged mainly by how many threats they can detect, or by independently demonstrated accuracy in the conditions where soldiers will rely on them?
Sources and evidence
- The US Army is putting money on AI drones that can find hidden threats and protect its Abrams tanks – Business Insider (12 November 2026, 16:06 UTC)
Independent WittyWires tracker for public updates about Anduril AI. Not affiliated with or endorsed by Anduril AI; this is not an official account.