Debain's broken OpenSSL claimed to be a cryptographically secure true random number generator (CSRNG). But it ended up being seeded with only 15 bits of entropy, so it failed in the true random part. Nevertheless, the pseudorandom number generator (CSPRNG, or as NIST calls it a DRBG) part of it still sorta worked (I don't recall if you could successfully seed it manually).
But regardless of how you were planning to use it, if an adversary has a backdoor in your PRNG/DRBG then it's not cryptographically secure (CS). That, and this Dual EC contraption is probably much slower than a conventional design.
