Arctic wargaming with AI: Part 2
Claude's NATO opts for aggressive deterrence as Perplexity's Russia manoeuvres for advantage
This is Part 2 of my AI-driven exercise, premised on a Russia-vs-NATO standoff during the US Presidential transition in December-January 2028-9. See my previous post for how we got here. In this edition we go straight into the action.
Events of 15.12.2028 – 04.01.2029
15 – 17 December:
Pro-Russian voices in Sweden, Ireland and Germany declare NATO aggression in the high north risks militarising Svalbard. Small titushki demos begin in these countries.
JEF HQ at Lossiemouth becomes 24/7 combined ops HQ, with direct links to Norwegian HQ at Bodø, Copenhagen, and the Icelandic Coast Guard. NATO general secretary and a UK defence minister stage a joint visit for the cameras.
RAF P‑8s and a Wedgetail from Lossiemouth and Norwegian P‑8s from Evenes surge into the Norwegian Sea/Barents approaches. There are no firm Russian submarine contacts.
The British Astute sub is routed towards the Barents Sea to form a forward defence
SNMG1 forms an Anti Submarine Warfare (ASW) screen around 72°N 5°E, between Jan Mayen and Norway
Russian influence operations alleging “Norway is militarising Svalbard” increase. The Duma passes a symbolic resolution annexing Svalbard.
18 – 22 December
The JEF Joint Commando Force is pictured exercising at Bardufoss in full Arctic kit. A visiting Norwegian minister comments that the JEF Commando force is “ready to move anywhere in the region at four hours’ notice”.
Micro-targeted Russian disinformation begins in Norwegian communities in Kirkenes, claiming their traditional way of life is being put at risk by militarisation. A peace campaign against the reservist alert stages a noisy demo in Oslo.
Russian troops in Karelia, Pskov and Belarus deploy for unplanned exercises – accompanied by an increased tempo of combat air exercise flights in the Leningrad Military District AO.
A Russian hack-and-leak operation releases “classified” NATO/JEF assessments warning of Russian preparations for major operations around Suwałki, Narva and Transnistria. Russia issues a diplomatic protest claiming these threats are fabricated by NATO.
Danish F-35s begin Combat Air Patrols (CAP) over Eastern Greenland, with NATO tanker support. Danish and Norwegian transport aircraft begin ferrying infantry, medium-range GBAD and mobile radar installations to Bluie East 3 and Jan Mayen Island.
The North Atlantic Council gives JEF HQ full authority to respond with force without consultation to any “Russian boots on the ground/Little Green Men” event in Svalbard, Jan Mayen or Kirkenes.
A 2x frigate Norwegian naval task force forms East of the Faroe Islands.
4x RAF Typhoons rotate into Keflavik, and RAF transport aircraft deliver stronger GBAD to the airbase. The Czech Gripens at Keflavik begin high-tempo CAP in the Iceland-Jan Mayen region.
(!) A Russian SSK is detected by Atlantic Bastion transiting South between Greenland and Iceland. JEF HQ decides to make no response.
(!) The Russian “research” vessel Yantar appears 30 nm West of Jan Mayen Island
Russian land forces on the Norwegian border become noticeably more active, with active radar and jamming operations and 2x Grisha class corvettes buzzing around the maritime demarcation lines, shadowed by Norwegian fast attack boats.
Norwegian F-35s begin regular CAP in Finnmark.
23 – 27 December
Norwegian Finnmark and Brigade Nord reserve call up is complete. Troops deploy to coastal, infrastructure and border defence positions. Forward units in Finnmark begin using drones right up to the edge of Russian airspace.
28 – 31 December
Russian rhetoric around the High North noticeably subsides. Most of the rhetoric is focused on the “oppression of Russian speakers” in Narva. Frontline Russian troops opposite Estonia are now sleeping in their vehicles, per invasion SOPs. Unattributable (i.e. “civilian”) Russian drone incursions are happening daily in the Narva AO.
The Norwegian TF (2x Frijdhof Nansen class FFGs) is joined by a British Type 45 destroyer and a Danish Iver Huitfeld Class FFG.
(!) Arctic Bastion picks up a Russian SSN transit between Greenland and Iceland. There are now two confirmed Russian subs SW of Iceland. The second sub is tracked by an RAF P-8 but lost.
1 – 3 January 2029
At a New Year celebration in Barentsburg, Russian miners clash with Norwegian police; two policemen are shot and injured. The suspects flee to the deserted Russian settlement of Pyramiden in a civilian RHIB.
(!) A Russian Duma delegation lands by unscheduled flight at Longyearbyen, together with Russian news crews and bloggers, is demanding an international inquiry into the incident.
Norway despatches a planeload of armed police to Svalbard via civilian charter flight.
The Norwegian coastguard vessel Svalbard is stationed off Pyramiden observing the movements of the fugitives and other unknown men who have gathered there.
There is intense propaganda in the Russian media against NATO aggression and Arctic militarisation.
Attributable Russian military drone incursions begin on the Russia-Norway border.
JEF is now staging a regular pulse of CAP patrols out of Evenes, Keflavik and Skrydstrup.
Here is a map showing the situation on 3 January 2029:
And here is the detailed picture of the incident on Svalbard:
How the NATO commander is thinking
At 20:00 3 January 2029 the JEF Commander reports to the North Atlantic Council (verbatim quote from the Claude output):
“I am establishing a posture of forward deterrence by detection and rapid reinforcement, backed by a potent ASW capability in the Norwegian Sea and pre-positioned cyber and air options.
I cannot be strong everywhere, but I can ensure that Russia cannot achieve surprise, cannot secure a fait accompli before we respond, and will face immediate and escalating military consequences at sea and in the air if they move.
The political messaging - to Moscow, to Washington, and to our own publics - is as important as the military dispositions.
I am acutely aware of two risks.
First, the Svalbard constraint: I cannot pre-position there but I must be able to respond within hours.
Second, the US factor: I am planning on the assumption that American forces will not fight, but American intelligence will flow.
If either assumption changes, I will adapt.”
How the Russian HQ is thinking
OK so now it’s time to reveal some of the workings of this game. In Turn Zero I fed the above SITREP into Perplexity (Best) and asked it, as the Russian player, to assess the effectiveness of the JEF’s response to Russia’s actions. Here is what it replied:
I note your measures. They are competent, energetic—and predictable. You have done two things well: you have built a fast headquarters, and you have tried to impose a single operational picture across the Norwegian Sea, Barents approaches, and Svalbard. Your standing naval group and P‑8 surge create activity, but activity is not the same as understanding, and your public posture creates a rhythm I can read and manipulate.
On your C2 activation (Lossiemouth)
A 24/7 combined room reduces your internal friction, but it also centralises your bias: you will fight for coherence, and I will feed you contradictions. The more you insist on one “truth” for the whole theatre, the more valuable ambiguity becomes as a weapon—especially where legal definitions and attribution drive your political will.
On your ISR surge (P‑8s, overhead, Astute)
Your P‑8s from Lossiemouth and Evenes increase the probability of contacts, not certainty, and you admit that yourself. That is the seam I exploit: I do not need to hide everything; I only need to keep you unsure about which contact matters, and when.
As for Astute in the Barents approaches: good. It forces caution, but it also tempts you into overconfidence—“we are inside their gates.” If you believe that, you will attribute routine Northern Fleet movements to “preparations,” and your own narrative will drift toward escalation. Then my diplomats will stand on Article 9 and ask Europe why you are militarising a demilitarised archipelago “for warlike purposes,” while Norway claims the treaty still allows minimal sovereign presence such as Coast Guard patrols.
On SNMG1 and the Norwegian Sea screen
You place SNMG1 as a visible shield, and you assume it gives you freedom of manoeuvre. It also gives me a convenient, lawful target for shadowing, and it pulls your attention to the Norwegian Sea—exactly where I can make you spend hours chasing signatures that do not matter.
Your group is strong at ASW and presence; it is not an answer to a low-signature fait accompli on land under civilian cover. Every hour your helicopters prosecute “possible submarines,” I gain decision-space elsewhere.
On Bardufoss readiness and Finnmark posture
You have armed your own reflex: Finnmark feels like the main event because it looks like the main event. Reservists, GBAD activation, CAPs—these are all loud signals, and they condition your staff to expect the decisive move on the mainland.
I am content with that.
On your communications campaign
You are right: attribution stiffens resolve. That is why, from my side, anything “influence” will be clean, boring, and deniable—lawyers, “concerned locals,” commercial disputes, safety incidents, and legal notes.
You have also tied yourself to defending Svalbard “under the treaty framework.” That is a double-edged sword because Article 9’s language on naval bases, fortifications, and “warlike purposes” is short and politically malleable; Russia has historically pushed a stricter reading than Norway, and I will exploit that gap to keep at least some European audiences uneasy about any rapid allied move onto the islands.
What I conclude by 4 January:
You have improved reaction time, but you still have an attribution problem: “boots on the ground” will not look like a brigade landing; it will look like disorder, police action, or protection of civilians until it is too late.
Your posture is observable and therefore gameable; your own readiness becomes my targeting data (timing, attention, and political messaging).
The decisive contest is no longer who has more ships and aircraft in the Norwegian Sea; it is whether your coalition can act fast under legal and narrative fog—on an archipelago where the demilitarisation clause is constantly contested in interpretation.
If you want my candid estimate: you have made it harder for me to win cheaply through NATO indecision alone. You have not yet made it impossible for me to win quickly through ambiguity.
Comment and Analysis
It may have been the way I prompted them, but Perplexity, as the Russian HQ, is now actively manoeuvring to get inside Claude’s OODA loop and conceiving the plan more holistically than Claude.
I will try to remedy this in the next turn. Perplexity/Russia seems to understand the legal complexities of the Svalbard situation very well - and also replicates Russia’s cognitive warfare doctrine well.
Claude has done the most aggressive things NATO could realistically do, but the situation on Svalbard is now critical - and that may not even be the centre of gravity of the Russian operation.
The next turn will probably be decisive from the point of view of deterring an Article V crisis. So watch this space and - as always - please share and subscribe if possible.