The Terminator

The fiction defines the nightmare. The table scores what reality has already built — each symptom checkable, each status sourced, the absences counted as honestly as the arrivals.

The symptoms

	symptom	the evidence	as of
✅	Loitering munitions that select and engage targets autonomously, used in real combatpresent	Ukraine's military intelligence (HUR) states Russia's V2U loitering munition can "autonomously search for and select targets using artificial intelligence"; first combat use February 2025 in Sumy region, 30–50 sorties per day by mid-May 2025, including a documented May 2025 strike by seven V2Us on vehicles and a market in Velykyi Burluk. The earlier candidate case — STM Kargu-2 in Libya, UN Panel of Experts report of March 2021 — remains contested; the V2U reporting is not. [source]	2025-06
✅	AI targeting-recommendation systems steering strikes in active warspresent	+972 Magazine reported in April 2024 that Israel's "Lavender" system marked roughly 37,000 people as strike candidates in Gaza, with human review described by sources as brief; "The Gospel" recommends structures. Separately, the Palantir-built Maven Smart System supports US target identification, and NATO signed a contract for its own version on 25 March 2025. Humans still authorize; machines now produce the target lists. [source]	2025-03
✅	State-funded programs to field autonomous weapon swarms at scalepresent	The Pentagon's Replicator initiative (announced August 2023) declared its goal of fielding thousands of attritable autonomous systems achieved in August 2025 — critics counted hundreds — and was folded into the Defense Autonomous Warfare Group in late 2025, which received $225.9M in FY2026 with up to $54.6B requested for FY2027 autonomy and drone programs (Defense One, May 2026). The funding line is real and growing. [source]	2026-05
✅	A legal vacuum: no binding international law governs autonomous weaponspresent	As of the March 2026 Geneva session, the UN CCW expert group has produced only a non-binding "rolling text"; the forum's consensus rule lets a small group of states block prohibition language. 156 states backed a UN General Assembly resolution and 42 CCW states declared readiness to negotiate, but the UN Secretary-General and ICRC's call for a binding instrument by end-2026 is unmet; the Seventh CCW Review Conference in November 2026 is the decisive test. [source]	2026-03
✅	Major-power doctrine permits human-out-of-the-loop lethal forcepresent	US DoD Directive 3000.09 (updated January 2023) does not prohibit weapons that select and engage targets without human intervention; it requires "appropriate levels of human judgment" and a senior review that the Deputy Secretary of Defense may waive in cases of "urgent military need." In May 2026, US senators publicly argued the directive is not keeping pace with fielded autonomy. No US, Russian, or Chinese policy bans full autonomy in conventional weapons. [source]	2026-05
◐	Autonomous kill-completion as mass-produced standard equipmentpartial	Ukrainian terminal-guidance modules (The Fourth Law's TFL-1) autonomously fly the final 400–500 meters through jamming, in systematic frontline use for 6–9 months; Vyriy Drone and The Fourth Law announced mass production in September 2025, with most frontline FPVs expected fitted within a year. Partial because in these systems a human still selects the target; the machine only finishes the kill after the radio link is severed. [source]	2025-12
—	AI holding nuclear launch authority — Skynet's actual jobabsent	No state delegates nuclear launch to a machine. The 2022 US Nuclear Posture Review commits to a human "in the loop" for all nuclear employment decisions; on 16 November 2024 Biden and Xi jointly affirmed that humans, not AI, must control nuclear weapons decisions — China's first such statement — and December 2024 US legislation codified safeguards against AI-initiated launch. This is the film's central premise, and it is affirmatively defended, not merely missing. [source]	2024-11
—	Self-improving military AI — weapons that upgrade their own capabilitiesabsent	No documented fielded weapon retrains or rewrites itself autonomously. The closest real practice, reported June 2025, is a human-run loop: Russia reportedly flies V2U sorties partly to collect data that engineers use to retrain the drone's AI. Recursive self-improvement remains a research agenda (an ICML 2026 workshop topic), not a deployed military capability; every fielded system's learning is done offline, by people. [source]	2026-06

The honest read

A high count of "present" here does not mean Skynet; it means the tactical half of the nightmare — machines picking and killing individual people, under no binding law — is documented battlefield fact as of 2025–2026. The film's two load-bearing elements, machine control of nuclear weapons and self-improving AI, are absent, and the first is actively defended by US policy and a 2024 US-China joint statement. The real barrier is not self-aware machines; it is the legal vacuum of symptom 4 — human judgment over lethal force is currently a waivable policy choice, and November 2026 in Geneva decides whether it becomes law. Base rate for calibration: automated defensive weapons (mines, close-in ship defenses, the Harpy anti-radar drone) have existed for decades without escalation to machine-directed war; what changed is that target selection itself moved into software, at scale, in two live wars.