Most crypto implementations are vulnerable, yes, but the percentage of vulnerable deployed crypto implementations as a fraction of all deployed implementations

Most crypto implementations are vulnerable, yes, but the percentage of vulnerable deployed crypto implementations as a fraction of all deployed implementations looks much better. Note that I'm talking crypto: the OpenSSL people seem quite able to make all sorts of mistakes, but it's almost never the crypto itself that is at fault. (Today's news: a workaround for Netscape's stupidities apparently allows attackers to downgrade to a worse protocol suite in certain cases. Stupid bug, but not crypto per se.) So yes, there are errors, but publishing something AES-encrypted and giving away, say, half the key doesn't tell us anything about those.