Proof of Reserves is a good start, but it needs to be coupled with Proof of Liabilities in order to be useful.

You have to be able to say:
1. sum of reserves >= sum of liabilities
2. there are no *other* liabilities

(1) is relatively easy, by providing a Merkle tree of reserves and a crowdsourced list where everyone can check that their own account is listed as a liability.

(2) is a hard problem. The CEO could obtain an off-the-proof/off-the-books line of credit and gamble it away, causing an additional liability not in the PoR/PoL proofs.

But there are lines of defense against even this. Creditors could refuse to extend credit to someone who has issued PoLs unless the credits appear in the PoL. Better yet, all corporate actions could flow through a contract that incorporates them into PoRs/PoLs.

Note that such practices are viral. I'm convinced that better accounting practices can originate in crypto and come to change the way we do business in general.

Another approach is to constrain, using crypto, what the custodian can do with the funds in the first place. This is what happens with DEXes (e.g. TraderJoe, Uniswap, etc) and FEXes (enclave.market).

It's easy to say "there are no other liabilities" if the custodian is circumscribed by code such that they cannot engage in extra-contractual activities.