FTX didn’t just fail; it exposed how a fast-growing exchange can become a fragile financial institution when basic controls are missing. By 2025, the best-run centralised exchanges treat their business less like a high-speed tech company and more like a regulated custody-and-brokerage hybrid: conservative treasury rules, strict separation of duties, and systems designed for “run risk” when customers all want their assets back at the same time. International regulators have also raised the baseline expectations for governance, custody, and conflicts of interest, pushing the sector toward more measurable standards.
One of the sharpest post-FTX shifts is the focus on custody design. Exchanges increasingly separate customer assets from the firm’s own operating funds, then enforce that separation with process controls: dedicated wallets, strict permissions, and independent approval paths for any movement of assets. The point is practical: if the business hits trouble, customers should not be competing with other creditors for assets that were never meant to be part of the firm’s balance sheet in the first place.
On the operational side, daily (or near real-time) reconciliation has become a core discipline rather than an audit-season exercise. Risk teams now expect a clean chain from customer liabilities (what users are owed) to on-chain balances and custodied holdings, with exceptions investigated immediately. That sounds obvious, but it is exactly the kind of boring, repeatable control that prevents “phantom balances” and hidden shortfalls from building up unnoticed.
Regulation is reinforcing this direction. In the EU, the Markets in Crypto-Assets Regulation (MiCA) sets a common regime for crypto-asset service providers and includes governance and organisational requirements that push firms toward clearer custody accountability. ESMA’s role in publishing registers and supervisory information is part of a broader attempt to make the market more legible for customers and authorities.
Segregation is not just a line in terms and conditions. In mature operations it usually means (a) wallet separation by purpose (customer custody vs treasury vs fee revenue), (b) access control policies that prevent single-person asset movement, and (c) an internal ledger that can be independently checked against wallet balances. If any of those layers is missing, segregation becomes a marketing claim rather than a control.
Another practical improvement is tighter treatment of “hot” vs “cold” holdings. Hot wallets exist for withdrawals and operational flows; cold storage is where the bulk sits with heavier controls. Post-FTX, many exchanges reduced the fraction held hot, added more automated limits, and introduced step-up checks (extra approvals) when thresholds are exceeded. That limits the damage from both internal mistakes and external compromise.
Finally, custody risk is now assessed like a system, not a single component. Exchanges increasingly run operational resilience exercises: what happens if a signing device fails, if a key employee is unavailable, if a chain halts, or if a major wallet provider has an incident. The “what if” work is unglamorous, but it is the sort of preparation that tends to matter when markets are moving fast.
FTX highlighted how quickly confidence collapses when customers suspect an exchange cannot meet withdrawals. In response, many exchanges strengthened liquidity management: they track withdrawal behaviour, maintain liquidity ladders (what can be turned into withdrawable assets within hours, days, weeks), and keep buffers in highly liquid instruments rather than stretching for yield. The goal is simple: survive a demand spike without selling long-dated or illiquid assets into a falling market.
Risk limits around lending and collateral have also tightened. Where exchanges offer margin, loans, or yield products, the stronger firms apply conservative haircuts to volatile collateral, set concentration limits (so one coin or one borrower cannot dominate risk), and run stress tests that assume large price gaps and thin liquidity. This is aligned with global standard-setter recommendations that emphasise robust risk management, conflicts controls, and market integrity measures for crypto-asset markets.
A key post-FTX lesson is that “solvency” is not only about assets; it is also about liabilities and contingent obligations. Exchanges are increasingly formal about what sits off-balance-sheet: guarantees, insurance commitments, affiliate exposures, and any obligations created by internal market-making. Better operations treat these as first-class risks with governance scrutiny, not as footnotes.
After FTX, proof of reserves (PoR) became widespread as a way for exchanges to demonstrate they hold assets corresponding to customer balances. Many PoR approaches rely on cryptographic techniques such as Merkle trees to let users verify inclusion of their balance in a liabilities snapshot, while the exchange publishes corresponding reserve addresses.
However, PoR is not automatically “proof of solvency”. A snapshot can show assets at one point in time while failing to capture borrowed funds, undisclosed debts, or liabilities elsewhere. That is why 2025 best practice is moving toward richer disclosures: repeatable PoR schedules, clearer scope statements (what is included and excluded), and—where possible—methods that reduce the risk of hiding liabilities while preserving customer privacy.
For readers, the practical takeaway is to treat PoR as one input, not a guarantee. It is most meaningful when paired with governance signals (independent audits, transparent risk policies, and credible supervision) and when the exchange publishes methodology details instead of only a headline reserve ratio.

FTX also exposed governance failures: blurred lines between related entities, weak oversight of treasury decisions, and internal control gaps. In 2025, higher-quality exchanges are formalising governance in ways that look familiar from traditional finance: independent risk and audit functions, documented policies for treasury and credit, and clear escalation paths when limits are breached. The spirit is to make risky decisions visible early, not after the damage is done.
Another major theme is conflict-of-interest management—especially where an exchange has affiliated trading entities, market-making desks, or token issuances. Stronger governance separates decision-making, restricts information flows, and requires explicit approvals for related-party activity. Global standard setters have stressed conflicts and market integrity as key concerns in crypto and digital asset markets, and those expectations have become harder for large firms to ignore. :contentReference[oaicite:7]{index=7}
Finally, supervision is becoming more concrete. In the EU, MiCA creates a harmonised framework for service providers and a clearer supervisory map, while in the UK the FCA has been advancing proposals for regulated stablecoin issuance and cryptoasset custody, signalling more detailed requirements for safeguarding and operational controls as the regime develops.
If you are assessing an exchange in 2025, start with the boring questions: who holds custody, how segregation is implemented, how frequently liabilities are reconciled, and whether the firm explains these controls plainly. A credible exchange can describe custody operations without resorting to vague reassurance, and it can be specific about what happens during high-withdrawal periods.
Next, look for limit discipline. Exchanges that take risk seriously disclose (at least at a high level) how they manage concentration, collateral haircuts, and liquidity buffers. They also tend to be cautious about promising yield, because yield usually implies someone is taking the other side of a risk trade.
Lastly, pay attention to regulatory posture and transparency cadence. Under regimes like MiCA and emerging UK rules for custody and stablecoin-related activities, the direction of travel is toward clearer governance, safeguarding expectations, and accountability. Firms that are already operating with those standards in mind are generally less likely to rely on improvisation when markets turn against them.