Not quite true...
IANAL but the text of RIPA Part III is very simple and easy to read and you don't need them to believe anything, all you need is a plausible enough message which is intelligible enough to an average human reader when decrypted.
One day, AI will be able to generate slightly different fully intelligible messages for all recipients based upon what appear to be identical ciphertexts, completely buggering any forensic methodologies for tagging messages without compromising every single sender. Each recipient would know what to mentally discard but each recipient would have to mentally discard different parts of the resulting message to obtain the same contextual meaning. This would poison the well so badly that even were a single witness to snitch, other recipients could claim plausible deniability all having received different messages in the first place!
This is the beauty of One Time Pad and a potential use case for homomorphic encryption, depending on the class of information being processed. You don't have to prove that the key you're using re-creates the true original message, only that it makes the protected information in question intelligible. Therefore, generating a set of decoy keys to create plausible dummy messages from ciphertext is actually a valid way to circumvent s.49 notices.
You can even have a situation where every witness supplies their (each very different messages) openly to the police with no legal risks. After all, if everyone delivers a slightly different end result, who is perjuring and who isn't? (answer: nobody) If all versions seem to somewhat fit the facts but are different enough in seemingly inconsequential ways, it isn't reliable evidence or testimony anymore.
This would leave the age old tactic of psychological intimidation as the only remaining weapon and encryption will be back where it started: A get out of jail free card for intelligent people to use at will.
Biting the hand that feeds IT
