The only parts of the report that give me pause are those that advance the claim that armed drones, specifically, can raise the risk of interstate conflict. Unlike most analysis on armed unmanned aerial vehicles which focuses solely on the counterterrorism context, the authors pay attention to implications for traditional war and escalation.
Their argument, as I understand it, is as follows. Armed drones are indeed different from other weapons systems, even in conventional settings. They are likely to be used more frequently for targeting or probing because the political cost of aircraft loss is lower compared to manned systems (the report refers to this as "low risk"). Furthermore, the authors claim that two miscalculation dangers are unique to armed drones: (1) misidentification of a manned aircraft as merely an armed drone might lead to an unintended attack on a pilot, which could then escalate; and (2) ambiguity on rules of engagement could lead to unintended escalation if a drone is shot down unexpectedly or if the response to a shootdown is unexpectedly severe.
Let's think about the two miscalculation dangers. Regarding (1), the authors worry that "Japan may misidentify a Chinese manned fighter as an advanced drone and fire on it, especially if the aircraft's radar signature is not sufficiently distinctive or if combat drones routinely fly over the disputed area." I do not disagree that this is a danger, even if I think adversaries not already at war would try to establish visual contact given the mishaps that have occurred in the past (also note that the report itself states that pilots are more responsive than UAVs to radio and pilot warnings). It's just that this argument provides just as much justification for limiting the use of surveillance drones as armed drones. If a manned aircraft can be mistaken for an armed drone on radar, it can also be mistaken for a surveillance drone. And if unarmed drones regularly fly over disputed areas and later a manned aircraft does, the same pattern-based risk exists. This undermines the authors' assertion that "though surveillance drones can be used to provide greater stability between countries by monitoring ceasefires or disputed borders, armed drones will have destabilizing consequences." If the misidentification logic is truly important, the authors should consider calling for export limits on surveillance aircraft as well, leaving monitoring to substitute systems.
Second, ambiguity about the "rules of engagement," according to the authors, could also lead to escalation (2): "how would states respond to an armed drone in what they contend is their sovereign airspace, and how would opposing sides respond to counter-drone tactics?" This is a more serious worry yet, again, it seems to apply to all unmanned vehicles, not just those which carry arms. The authors are right to restrict their focus to disputed territory and airspace: where boundaries and identification zones are clear, I think government expectations already converge on when aircraft of any kind would be shot down if detected. But given the uncertainty of responses above disputed zones, challengers with drones do indeed have a new tool with which to probe the adversary or to change "facts on the ground." (Targeting is a different story given the existence of other unmanned systems like missiles and artillery.) Here we bring in the low-cost argument: drones may be employed by casualty-sensitive territorial challengers where manned aircraft would not be, presumably to to show/test resolve or to further claims about de facto control.
The problem for the report is that, in one of the disputes it draws several examples from, the deployment of unarmed drones has escalated tensions and highlighted a disagreement about rules of engagement. Surveillance drones were at the core of a flare-up between Japan and China in late 2013, with Japan afterwards stating that it would shoot down unresponsive UAVs in its airspace and China claiming that this would be an act of war.
Furthermore, observers question that Japanese stance, asking, "how could an unarmed surveillance drone meet the SDF’s rigorous criteria [involving self-defense] to allow shots to be fired?" So in this particular case, arming drones might actually bring more certainty about likely responses by making the defender's threat to shoot more credible and by rendering any subsequent Chinese claim of Japanese aggression less legitimate.
In conclusion, I am only questioning part of the report. I do not believe the strong assertion that "current practice repeatedly demonstrates that [weaponized] drones make militarized disputes more likely," both because we have little empirical basis for this claim and because surveillance drones may pose the same dangers. Unmanned aircraft might indeed increase the likelihood of interstate war through miscalculation. Yet the only logics that survive to buttress this claim apply at least equally well to unarmed UAVs. This fits awkwardly into an otherwise valuable report which seeks to single out armed drones as especially dangerous.