When securely distributing a symmetric key over an untrusted network, a common approach is hybrid cryptography: use asymmetric cryptography to protect the symmetric key, then use the symmetric key for bulk encryption. To ensure only Employee B can recover the symmetric key, Employee A encrypts (wraps) that symmetric key using Employee B's public key. Because only Employee B should possess the matching private key, only B can decrypt the wrapped symmetric key. This is the same principle used in TLS key exchange (in older RSA key transport) and in secure email: encrypt the session key to the recipient's public key. Encrypting the symmetric key with Employee A's private key would not provide confidentiality---anyone with A's public key could reverse it, and it functions more like a signature than encryption. Employee B's private key should never be shared and is used only by B to decrypt. Therefore, for confidentiality of the shared symmetric key, the correct encryption key is Employee B's public key.