Two inequivalent definitions of a Markov blanket are in use, and a result shown under one does not carry over to the other. One restricts which coordinates drive which in the equations of motion (a dynamical condition); the other requires inside and outside to be independent given the blanket in the steady-state density (a statistical condition). Neither implies the other: a sparsely coupled system can lack the conditional independence, and the reverse also occurs. Adding to the dynamical condition the assumption that solenoidal coupling between blanket, inside and outside vanishes collapses the four sets into a chain, internal states influenced only by active ones and external only by sensory ones, and in that chain calling one set sensory and the other active is no longer meaningful.

Evidence 3 passages

groundsA Technical Critique of Some Parts of the Free Energy Principle, Martin Biehl, Felix A. Pollock, Ryota Kanai
Neither Condition 1 (the vector field dependency structure) nor Condition 2 (conditional independence in the ergodic distribution) imply the other
groundsA Technical Critique of Some Parts of the Free Energy Principle, Martin Biehl, Felix A. Pollock, Ryota Kanai
we reveal that various definitions of the "Markov blanket" proposed in different works are not equivalent
groundsA Technical Critique of Some Parts of the Free Energy Principle, Martin Biehl, Felix A. Pollock, Ryota Kanai
it is questionable whether the $s$ and $a$ coordinates can be meaningfully interpreted, respectively, as sensory inputs to the internal coordinates or their boundary-mediated influence on the external coordinates