Threat assessments usually rank Russian options by capability and target value. Also important seems the deniability calculus: what Moscow can plausibly deny, what it wants us to suspect, and the span of attention of each threat.
The catalogue problem
The Atlantic Council recently published a useful survey of where and how Putin could expand his war in Europe beyond Ukraine. Region by region, from the Baltic states to the Black Sea, its authors list the instruments: sabotage, cyber operations, disinformation, GPS jamming, drone incursions, pressure on critical infrastructure, and the possibility of a limited military crisis designed to test NATO’s cohesion.
The catalogue is sound. Remains to rank the options. Attribution is largely treated as settled: incidents are “linked to Russia”, prosecutors “have concluded”, authorities “have held Russia responsible”. That is how attribution looks in retrospect, in the cases that were solved. It is not how it may look in real time, when a cable is cut and three explanations compete.
One variable is deniability, and that it can be used systematically: one important filter (but of course not the only one) for judging which Russian attacks are probable, which are possible, and which are unlikely because they may be difficult to deny.
Coercion needs a return address
Start from a tension that classical deterrence theory identified long ago: coercion requires attribution. A threat only compels if the target knows whom to yield to. A perfectly deniable act therefore may limit pressure; it damages, but it does not necessarily bend anyone’s will.
The reverse is also true. A fully attributable act, such as a state aircraft violating allied airspace, carries maximum signal but forfeits all protection. It invites a collective response and may create the alliance consensus that ambiguity is designed to prevent.
Russian practice under the threshold of war seems ofte to operate in the middle between these poles. Implausible deniability: acts deniable enough to block a legal or alliance-wide finding of state responsibility, yet transparent enough that the intended audience understands the message. “You know we can” as a message may be more useful to Moscow than “we did”, because it plants a lever that can be activated later.
Sabotage through recruited proxies is perhaps the cheapest available demonstration of both capability and will: high threat effect, low escalation risk, and a legal chain that stops at a (paid) intermediary. Nuclear signalling, by contrast, is expensive and has been devalued by overuse; the credibility-fatigue literature documents how repeated nuclear threats lost effect as crossed “red lines” produced no consequences. The deniable middle is where the remaining coercive value may continue to live.
Three tiers of deniability
Applied to the Atlantic Council’s catalogue, the filter sorts the instruments into three tiers.
High deniability, high probability. Proxy sabotage, cyber operations behind criminal or “patriotic” fronts, disinformation, political and economic subversion of the Moldovan type, drifting sea mines that can be blamed on the war itself, and damage to undersea infrastructure by shadow-fleet vessels dragging anchors. The anchor “accident” deserves particular attention because it shows the mechanism in pure form: a civilian ship, a commercial pretext, a physical explanation that is individually plausible. Expect continued and expanded use of this tier.
Conditional deniability, self-limiting probability. Drone incursions retain deniability at the moment of incident but lose it under repetition over protected sites. GPS jamming is the interesting hybrid: the emitter can be located by direction-finding, so origin is not deniable at all, while intent still is (“exercises”, “self-protection”). Assassination plots are deniable in execution but politically undeniable once unravelled; the accumulated record makes each new denial less credible than the last. False-flag operations sit at the extreme: engineered for deniability, yet carrying the highest exposure cost in the entire catalogue, since an exposed false flag may be worse than an admitted attack.
No deniability, low probability short of a decision for war. Manned airspace violations are signalling, not covert action. A limited military move against allied territory, the Baltic or Gotland scenario, is overt coercion by definition. It removes, in one stroke, the ambiguity on which the whole campaign depends. But extreme caution in determining probability is warranted as the full scale invasion of Ukraine 2022 illustrates.
Deniability depreciates
The filter has a second, dynamic component: deniability is a wasting asset. Each incident in a category consumes credibility. One damaged cable is an accident; the fifth is a pattern. Analysts of the shadow fleet have argued that only a systematic approach can deal with it, which is another way of saying that the accident narrative has already depreciated below usefulness.
Two curves are racing here. On one side, attribution accumulates: forensics improve, prosecutions conclude, patterns harden. On the other, the attention span is getting shorter: publics and governments absorb each incident into the normal picture, and only sudden or severe events still break through the noise. Which curve wins determines Moscow’s next move within a category. When deniability in one instrument class is exhausted while habituation is high, the rational choice is to find a new way. Early detection of new categories is key.
What the filter implies for policy
Some thoughts:
In any Baltic contingency, this means that attention has to remain also on the high-deniability tier directed at infrastructure, logistics and decision-making
Second, the Western counter needs to be to look for patterns. They require graduated responses attached to the pattern rather than to any single incident. Publishing attribution methodology in advance may raise the cost to Russia. The difficulty should be stated plainly: alliance consensus procedures are built to judge incidents, not patterns.
Third, the signalling half of the calculus should not be forgotten. Non-attributability is a problem which needs to be acknowledged: to state that the pattern is understood, priced in, and answered by hardening rather than alarm. Depriving the campaign of its psychological effect is important.
None of this makes the catalogue of Russian options shorter. It should make them more understandable. The question for the next incident is not only “was it Russia?” but “which tier was this, how much deniability did it spend, and what does that reveal about the choices Moscow believes it has left?”
Lars-Erik Lundin is a former EU Ambassador and Swedish diplomat.