Some background on military applications of the magnetrine effect, robotics, and AI, some of it sourced from Stålenhag's books:
The military applications of the magnetrine technology are plentiful – on paper. After many weapons projects, the strenghts and weaknesses of the technology have become manifest. First out were the Soviets, as Mikhail Vorobyev had discovered the effect back in 1943. The Second World War was over before it could be applied to weapons (it was originally a spin-off find while researching long-distance missile guidance systems), but during the first years of the Cold War, Soviet scientists experimented with magnetrine technology and armored vehicles. After Soviet scientist Vladimir Degtaryev defected to the West in 1951, the US got access to the technology, and soon several other Western countries.
One obvious area of research was armored vehicles. Magnetrine hover tanks proved to be problematic, though. The limitations of armor-to-lift ratio made the tanks too lightly armored, and coupled with the raised silhouette, hover tanks proved to be unsuitable for the modern battlefield. The area where magnetrine technology proved useful was in supply and transport. Magnetrine transports filled the niche between wheeled trucks and helicopters, combining capacity with all terrain access. Magnetrine cargo and troop transports became part of frontline units, as well as lightly armored reconnaisance vehicles. As the magnetrine tech worked best in the northern hemisphere, it couldn't be utilised in the Vietnam War or in many other conflicts in the Third World, and thus saw limited combat.
That limitation made magnetrine tech less useful for the world's navies. Still, magnetrine aircraft carriers were built by the US, Soviet Union, Great Britain and France. The technology was also useful for ship-to-shore transports. While traditional jet aircraft proved more versatile, breakthroughs in the early 1990's made fast reconaissance drones possible. If it hadn't been for the magnetic pole shift in 2001, magnetrine tech would have played a major role in the new generation of armaments.
When the self-balancing issue was solved by Iwasaki in the late 1960's, military self-balancing autonomous systems – or robots – soon saw the light of day. Classic, humanoid robots proved to be inferior on the battlefield, as they simply couldn't react and adapt like human soldiers. Instead, they were useful for sentry duty and load-carrying. Larger robots were more or less autonomous, mostly patrolling border areas, or remote-controlled by human operators, either with line-of-sight "control glove" systems, or via satellite uplink from control centers. As artificial intelligence improved, robots became more independent, like the Swedish-built ABM100. The Prague Convention of 1996, signed by all the major powers, as well as many smaller, industrialised nations, limited the use of autonomous robots in warfare. The fear of a singularity event, where AI systems would turn rogue, made it mandatory to install killswitches in all military robots.
Areas where military robots have proved valuable assets are mine and IED clearing, hazardous environment combat, urban combat support, load carrying, and as heavy weapons support platforms. The latter are quadroped robots armed with Gatling guns, guided missiles, recoilless rifles, and/or mortars, making them formidable assets in modern combat. The presence of robots on the battlefield has prompted the introduction of various EMP delivery systems, like mines, missiles and grenades. To counter this, the more modern robots have hardening to counter EMP and other countermeasures.
Problems with military robots have usually been associated with poor or rogue AI. The Baikal Wars in the 1970's saw Chinese "robot wave" tactics fail due to weak command and control. While the reasons behind the AI pogroms in Russia in the early 1990's remain unexplained, reports claim that some military robots were infected with viruses and turned on their human masters, while others refused to serve as soldiers, instead forming a hippie-like culture. As for Sweden, secret research conducted at the FOA facility at Sätra on Munsö include military robots. Despite safety measures, some of the robots are known to have escaped. It is possible that their AI is a little too good, as there are rumours that not all of them have been rounded up. Swedish Army AMAT-2 quadroped robots, painted in standard army splinter camouflage pattern, have been deployed from time to time in order to deal with rogue robots, strange beasts, and from 1995 onwards infected robots. The AMAT-2 and its predecessor, AMAT-1, have been successful exports, serving in the armies of e.g. Norway, Austria, Brazil, South Africa and Malaysia.