Automated Target Recognition (ATR) moves target recognition from a human operator onto the platform itself. A munition navigating this way is immune to datalink or GPS jamming — precisely the layer classic electronic warfare targets (see Chapter 4.2).
At larger scale this points towards swarms: coordinated employment of multiple autonomous platforms that together can saturate a defence chain faster than a human can assess and decide.
The next step is autonomous air-combat AI. In DARPA's ACE programme, AI flew the X-62A VISTA — a modified F-16 — in a real 2023–2024 dogfight against a human-piloted F-16. Its precursor, AlphaDogfight (2020), already saw an AI agent defeat a human pilot in simulation. Target acquisition and engagement at high altitude and speed are no longer a human monopoly.
AI also accelerates learning itself. Ukraine's OCHI system pools feeds from over 15,000 drone crews — roughly 2 million hours (~228 years) of footage since 2022 — on which AI models are trained; AI guidance raised FPV strike effectiveness from 30–50% to about 80%. Palantir's Maven system underwent 50+ rounds of improvement in its first ten months supporting Ukraine. This days-long feedback loop has no counterpart in the Western doctrine or procurement cycle (Lowy/Ryan, 2026) — linking this chapter to the adaptation cycle in Chapter 4.2.