It is perhaps ironic to accuse a purely capitalist scheme of having the disadvantages of central panning, but if the few Cyc officers with fiduciary responsibilities cannot understand a new idea that requires access to their data, the idea is now unlikely to be tested when those with the idea must pay commercial rates. Sometimes breakthroughs come from unexpected directions and from groups without the money to buy that access.
Market segmentation solves this in theory but only if you have viable segmentation scheme. Perhaps confinement as in the Derwent solution may allow this. The researcher’s program has unrestricted access to the data but there is a narrow hole thru which the researcher communicates with his program. Think gdb output which has not changed much since it was via the 10 cps Teletype.