I want to make this clear from the start, what you are about to read is speculative at best. This has been put together based upon what should be best practice for any automated defense platform of this type. For those who don't know what Perimeter is, it is the rumored automated nuclear counter-attack by Russia.
The following is how Wikipedia describes the system, based upon what little public information is available:
In 1967, the Soviet Union first attempted to create a system, called "Signal", which they could use to create 30 premade orders from their headquarters to the missile units. Although the system still was not completely automatic, their intent was no different.
In the early 1990s, several former high-ranking members of the Soviet military and the Central Committee of the Communist Party, in a series of interviews to the American defense contractor BDM, admitted the existence of the Dead Hand, making somewhat contradictory statements concerning its deployment.
Colonel General Varfolomey Korobushin, former Deputy Chief of Staff of Strategic Rocket Forces, in 1992 said that the Russians had a system, to be activated only during a crisis, that would automatically launch all missiles, triggered by a combination of light, radioactivity and overpressure, even if every nuclear-command center and all leadership were destroyed.
Colonel General Andrian Danilevich, Assistant for Doctrine and Strategy to the Chief of the General Staff from 1984–90, stated in 1992 that the Dead Hand had been contemplated, but that the Soviets considered automatic-trigger systems too dangerous. Furthermore, such systems became unnecessary with the advent of efficient early-warning systems and increased missile readiness, so the idea had been rejected.
In 1993 Vitaly Katayev, Senior Advisor to the Chairman of the Defense Industry Department of the Central Committee of the Communist Party in 1967–85, responsible for strategic arms and defense policy, arms control negotiations and military doctrine, confirmed that the Dead Hand had been "definitely operational" by the early 1980s. According to Katayev, it was not completely automatic but was intended to be activated manually during a threatening crisis. It was to be triggered by numerous sensors sensitive to light, seismic shock, radiation or atmospheric density.
Although both Katayev and Korobushin claimed that the mechanism had already been deployed, Viktor Surikov, Deputy Director of the Central Scientific Research Institute for General Machine Building (TsNIIMash) in 1976–92, confirmed in 1993 that the Soviets had designed the automatic launch system with seismic, light and radiation sensors, but said that the design had been ultimately rejected by Marshal Sergey Akhromeyev on advice of Korobushin and never materialized.
'The dead-hand system he [Dr. Blair] describes today takes this defensive trend to its logical, if chilling, conclusion. The automated system in theory would allow Moscow to respond to a Western attack even if top military commanders had been killed and the capital incinerated.
The heart of the system is said to lie in deep underground bunkers south of Moscow and at backup locations. In a crisis, military officials would send a coded message to the bunkers, switching on the dead hand. If nearby ground-level sensors detected a nuclear attack on Moscow, and if a break was detected in communications links with top military commanders, the system would send low-frequency signals over underground antennas to special rockets.
Flying high over missile fields and other military sites, these rockets in turn would broadcast attack orders to missiles, bombers and, via radio relays, submarines at sea. Contrary to some Western beliefs, Dr. Blair says, many of Russia's nuclear-armed missiles in underground silos and on mobile launchers can be fired automatically.'
The communication missile would work similarly to the US Emergency Rocket Communications System (ERCS).
However, more recent sources indicate the system was semi-automatic. In a 2007 article, Ron Rosenbaum quotes Blair as saying that Dead Hand is "designed to ensure semi-automatic retaliation to a decapitating strike." Rosenbaum writes, "Of course, there's a world of difference between a 'semi-automatic' doomsday device and the totally automatic—beyond human control—doomsday device."
David E. Hoffman wrote on the semi-automatic nature of Dead Hand:
And they [the Soviets] thought that they could help those leaders by creating an alternative system so that the leader could just press a button that would say: I delegate this to somebody else. I don't know if there are missiles coming or not. Somebody else decide.
If that were the case, he [the Soviet leader] would flip on a system that would send a signal to a deep underground bunker in the shape of a globe where three duty officers sat. If there were real missiles and the Kremlin were hit and the Soviet leadership was wiped out, which is what they feared, those three guys in that deep underground bunker would have to decide whether to launch very small command rockets that would take off, fly across the huge vast territory of the Soviet Union and launch all their remaining missiles.
Now, the Soviets had once thought about creating a fully automatic system. Sort of a machine, a doomsday machine, that would launch without any human action at all. When they drew that blueprint up and looked at it, they thought, you know, this is absolutely crazy.
So, all this would tend to suggest that based on the input of a number of sensors that detect the aftermath of a nuclear strike, Perimeter would launch a counter-offensive.
To me, that just sounds dumb.
Let me explain this, Russia much like the US has strong Artificial Intelligence, it has big data capability, remote sensing, remote tracking, etc, etc. In all, it is about on par with US capability in most key areas. Russia is a very practical minded fighting force, whereas the US has a tendency for big budget drama and the 'unexpected, unconventional, sci-fi' nonsense. As such, despite the disparity in wealth, Russia is probably the more capable of the two.
If you were to build a system like Perimeter, it would not be an after-the-fact mechanism, nor would it be dumb. It would be a warfare planning system running continuous monte carlo simulations based upon real-time information. A complex system monitoring millions, perhaps even billions of data points that were calculating the optimal response for the current state of global play. Not only that, but such a control system would, to the best of its ability, provide in-flight updates and defense support to all ICBMs and warheads. Basically, if a device could add anything to the fight, it would be co-opted, driven by the AI and as each device was destroyed a new optimal strategy would be deployed. No one has time for a network of humans trying to coordinate things whilst under fire.
To anyone in the world of Data Science, this should seem reasonable. Look at what we can do with Twitter feeds, open government data, etc. As such, if Perimeter exists at all, it is obviously triggered by a threshold based upon a trajectory of strategic failure. That is, if the system concludes that it is likely to placed in a position of strategic failure, it will launch and take its chances.
Not really. From a military perspective, the system has already determined that a post-apocalyptic outcome is preferable to the current state of play or the projection of it. Unlike government, whose fears you can put to rest, or pull the wool over their eyes an ever watchful IT system does not hear your excuses. It doesn't care, it just reads data, processes it and takes action.
Is a nuclear apocalypse preferable to surrender? In the case of a functioning nuclear deterrent, the answer is yes. A nuclear deterrent is nothing if you cannot press the button, or if you ignore computer screens telling you the other side has launched a pre-emptive attack.
Would the Kremlin know about this? Would Putin?
The truth is probably not. They don't need to know how things work, just that they do and there are red lines that should not be crossed. The only indication they may be given is that they need to push back on a move by the US, NATO or other party.
If this is the case, then the warfare planning AI currently determines the presence of the US, NATO, etc., around its borders not to be a viable threat. That said, Russia has issued warnings in key areas and these should be heeded. There is no reasoning with an AI once it has been programmed.
I would be wary of playing war games around the borders of Russia. Military posturing is just that, posturing, but an AI does not respond to posturing. It is just setting flags in memory. Whilst I can appreciate the need to make Eastern EU nations feel a sense of security, by sending signals, it serves no other practical military function.
As a species, I think we all know by now that everyone has armies, that there are weapons and we can do serious damage to each other. I don't think we actually need to see these things on the ground near anyone's borders.
They have been told where they stand and that should be enough.