The Kalashnikov family tree
Something like a hundred million rifles descend from one 1947 design, and almost none of them are what the word AK-47 actually names. Sorting out which is which turns out to be a lesson in how a design gets copied.
Start with a correction. The rifle almost everyone pictures when they hear AK-47 is not an AK-47. It is an AKM, or a licensed copy of one, or a copy of a copy. The genuine article (the milled-receiver rifle the Soviet Army adopted in 1949) was made for about ten years and replaced in 1959. Every stamped-receiver Kalashnikov made since is its successor rather than the thing itself.
That confusion is worth untangling, because the pattern of how this design spread is more interesting than the design.
Three receivers in twelve years
Mikhail Kalashnikov’s rifle won a Soviet trial in 1947 against entries from Bulkin and Dementyev, and its first production receiver was a stamped sheet-metal shell with a milled insert riveted in. Soviet factories could not hold the tolerances, rejection rates were unacceptable, and the design was changed to a receiver milled from a solid billet: stronger, much slower to make, and about half a kilogram heavier.
The stamping only came back in 1959, with the AKM, once the industry could do it. That rifle is lighter, has a different rear sight, a slanted muzzle brake and a stamped receiver, and it is the one that went everywhere. The AK-47 entry on this site carries the 1949 rifle’s figures and says so; its production total is recorded as an estimate because no manufacturer has ever published a count and the widely repeated seventy-five million is a survey figure.
Licensed, copied, and neither
The interesting part is what happened next, and the three different legal and technical relationships involved are why this site records lineage as typed edges rather than a single “based on” field.
The Type 56 is licensed production. The Soviet Union transferred drawings and tooling to China in the 1950s, and Norinco built the rifle to them. It is the same rifle made by someone else with permission, and the differences (a folding spike bayonet, a fully enclosed front sight hood) are deliberate local changes rather than errors.
The IWI Galil is derived from it. Israel came out of 1967 having decided the FN FAL was too long and too sensitive to sand, ran a trial, and adopted a rifle whose action is a Kalashnikov’s (long-stroke piston, rotating bolt, same basic geometry), licensed from Valmet’s Finnish version rather than copied from a Soviet one. Everything above the receiver is different: the stock folds sideways, the charging handle is bent up so either hand can work it, there is a second safety lever above the grip, and the whole thing is chambered for a NATO cartridge.
And then there is everything that is neither: Romanian, Bulgarian, Egyptian, Iraqi, Yugoslav and Pakistani production of varying legality and varying quality, plus an enormous commercial trade in receivers and parts kits. This site records what it can attribute and does not guess at the rest.
What made it copyable
The usual explanation is that the design is simple. That is half right and misses the more useful half.
The Kalashnikov’s tolerances are generous: clearances large enough that dirt, ice and badly made ammunition rarely stop the cycle. The same clearances are why it has never been an accurate rifle by target standards. Both halves of that trade were deliberate, and the important consequence is industrial rather than tactical: a rifle designed to work with loose tolerances can be made by a factory that cannot hold tight ones.
That is what a country with a new arms industry needs, and it is why the design propagated through exactly those countries. Compare the M16, which is a more accurate rifle made to closer tolerances, and whose licensed production went to established manufacturers in Singapore, Canada and Belgium rather than to new ones.
Reading the lineage graph
Every entry on this site that descends from another carries a typed edge saying how, and those edges have
to resolve to a real entry before the build will pass. That constraint has a visible consequence: the
Type 56 could only record its licensedProductionOf edge once the
AK-47 existed here as a page for the edge to point at.
The tree is therefore incomplete, and visibly so. The AKM, the AK-74, the RPK and the dozen national variants are not here yet, and until they are, the graph on the AK-47’s page shows two edges rather than fifty. That is an accurate picture of what this catalogue currently knows, which is the only kind of picture it is willing to draw.