Let's have a go, a flawed attempt though it may be.
A dog trained to protect its master and/or its master's intetests may end up suffering for its 'duty', at times, but normally the suffering is for the sake of the whole dog-man 'pack' - from which membership of, the dog enjoys reciprocal benefits, some duty of care. Whether or not the dog understands this, and fully appreciates it, as its ancestral pack-mentality is manipulated by the human (see "service dogs", including Diesel
, but with a contentious upper-end
to the analogy), only the human handler(s) truly having the capacity to know for what stakes the dog is playing the game, and thusly lies the incumbant responsibilities.
It is analagous (if not identical!) to a military command knowing best to what objective troops are deployed, and the risks they may endure, and knowing that the subordinates may not (perhaps should not
, at times) know the full import of their missions, but must trust that they are not being deliberately used for no net benefit.
In dog-fighting, the trust developed may be considered lop-sided. The dog's utility is not repaid by care to prevent needless risks. Their only purpose is to court
mortal risks, expendably so, and enter situations of their handler's engineering that need never ever have occured, but for an agreement between rival handlers in a relationship that trumps the previously expected 'our pack' bond. The fate of a winning dog is only to compete again, again and again, until not
successful; the fate of an (eventually) unsuccessful dog is rarely a reward of a happy retirement (which may not even be possible for one undefeated, without extreme care and effort). For the dog, there is no wise pack-leader to trust, however much pack-trust they 'feel'.
Counter-arguments to this include that such dogs only live
due to their potential to be sent into combat, and a life (however terribly endured in parts) may be considered better than never having a life in the first place (debatable), or an otherwise truncated one (differently debatable). Also that the pack (human plus succession of expendable, and often expended, dogs) may indeed benefit from the deliberate sacrifice of the canine party, by creating a better life for the human party by reputation and earnings1
, to advance (or maintain level-ish against a relative expectation of degradement) the standard of living of the revolving-door-pack, over time. Assuming that the betterment/sustenance of the (obviously dedicated and successfully ruthless) human trainer does not correlate with a zero or negative welfare gain on the part of the trainees. Still, even those caveated caveats may well be explored.
I suspect similar arguments surrounding slavery vs servitude vs employment (at various points on their much-overlapping spectra effective benificences from above) are applicable, but I couldn't quote you any chapter or verse from the extensive philosophical masters who have surely already covered every inch of this.1
This is also where the war-time bomb-dogs theoretically come in to the discussion at a meta-pack level.