I used to believe that turning off withdrawal rights on a trading connection was enough to keep my crypto safe. After watching what unfolded with one popular options platform earlier this month, that assumption feels dangerously incomplete. Trading permission by itself can rearrange holdings, rack up fees, open leveraged positions, and quietly shrink the value of an account long before any coin ever leaves the exchange.
How a Single Compromised Connection Exposed the Real Risk
The incident began outside any major exchange. An attacker first compromised a technology vendor’s messaging account, then posed as a trusted contact and sent a fake meeting invitation. That message led to an engineer’s computer being taken over. From there the intruder moved deeper into the company’s internal systems. Once inside, the platform that many traders used to monitor and execute options and derivatives strategies suddenly became a risk vector for every connected account.
Within hours the company detected unusual activity, shut down services, and asked partner exchanges to revoke every linked credential. Two large venues quickly disabled the associated keys and told customers to manage positions directly on the exchange itself. The company later reported that twenty-three users had suffered combined losses just under eighty-two thousand dollars and that full reimbursement reached those accounts within seventy-two hours. Yet the larger lesson sits far beyond the dollar figure.
Disabling withdrawals stops one exit route. It does nothing to stop a credential that can still buy, sell, open futures, or accumulate fees. I’ve found that most retail traders focus almost exclusively on the withdrawal toggle and treat trading rights as relatively harmless. That mindset leaves them exposed the moment the outside software is compromised or simply misbehaves.
What an API Key Actually Allows
An API key is simply a set of credentials an exchange issues so that external software can act on your behalf without logging in each time. You choose the permissions when you create it. Most platforms let you grant trade rights while denying any ability to move funds off the exchange. That separation looks clean on paper. In practice it still hands significant power to whatever system holds the key.
With trade permission alone a bot can:
- Buy assets that immediately drop in price
- Sell holdings before a rally begins
- Open leveraged positions that face liquidation
- Generate repeated trading fees that erode the balance
- Leave open orders that fill at unfavorable levels after the connection dies
None of those actions require a withdrawal. The coins stay inside the exchange wallet the entire time, yet the account’s net value can still fall sharply. Perhaps the most interesting aspect is how quickly volume can multiply the damage. A modest deposit can support dozens of round-trip trades; each one carries spreads and fees. Over a short window the total cost can exceed any reasonable expectation.
Exchanges themselves draw a clear line. One major platform’s documentation notes that a third-party trading service needs order permissions but almost never needs the ability to withdraw funds. Another separates “Trade” from “Transfer” in its permission matrix. The architecture exists. Traders simply need to treat the remaining rights with the same caution they already give to withdrawals.
The Chain of Events That Made the Risk Visible
According to the company’s own timeline, the attacker first used a compromised vendor contact account on a popular messaging app. Posing as that contact, the attacker sent a Microsoft Teams invitation that the engineer accepted. The machine was compromised. The intruder lingered, then leveraged the engineer’s privileges against the internal network on the morning of the fourth. Detection followed quickly. Services went into maintenance mode and every connected exchange was asked to suspend the relevant keys.
One exchange publicly confirmed it had revoked all related credentials and reminded users to review open positions and orders on its own interface. Another venue issued a similar notice. A prominent industry figure went further, urging anyone using the service on other platforms to revoke those keys as well and warning against storing exchange credentials with third-party tools in the first place.
The company later acknowledged that IP whitelisting had not been mandatory for the API connections. Accounts that suffered losses lacked that extra restriction. In its public statement the firm accepted responsibility for the oversight and listed stronger access controls, device security, and internal key management among the changes already under way. By the eighth it reported that all twenty-three identified users had been reimbursed for the full eighty-one thousand eight hundred ninety-seven dollars and that investigation and hardening work continued.
I keep coming back to one practical detail: even after the keys were revoked, any orders already accepted by the exchange remained live. Positions already opened continued to mark to market. A dead-man switch on some venues can cancel resting orders when the client stops sending heartbeats, yet that feature does not unwind fills that already happened. Customers still had to log in directly and decide what to do with whatever the bot had left behind.
Earlier Incidents Showed the Same Pattern
This was not the first time trading credentials held by an outside service created problems. In 2022 a different bot platform faced a wave of reports about unauthorized activity inside connected exchange accounts. One user described thousands of unwanted trades in a low-liquidity token that produced material losses. Initial explanations conflicted. Later the platform acknowledged the validity of keys that had circulated after earlier denials. Exchanges responded by refreshing credentials and adding monitoring. The common thread remains clear: once a key with trade rights is compromised, the exchange treats the resulting orders as legitimate.
A withdrawal history that shows zero outbound transfers can still hide a damaged portfolio. Different assets sit in the wallet. Fee balances have grown. Margin positions may sit closer to liquidation. The only reliable record lives on the exchange itself, not on a third-party dashboard that may have stopped updating the moment the connection failed.
Why Trading Permission Alone Is Enough to Hurt
Consider a simple example. Suppose a bot receives a key that can trade but cannot withdraw. An attacker or a faulty strategy starts buying an asset that is already rolling over. Each purchase locks in a higher average cost. Later the same key sells at a lower price. Spreads and maker or taker fees come out of the balance on every leg. After a few dozen cycles the account holds the same or even more units of the token yet is worth substantially less in dollar terms. No coin ever left the exchange. The damage is real.
Leverage multiplies the effect. If the credential can open futures or margin positions, a series of wrong-way trades can trigger liquidations. The exchange simply follows the rules attached to the account. It has no way to know that the orders no longer reflect the owner’s intent. In my experience that gap is where most of the quiet losses occur.
Another subtle point concerns allocation limits. Many bot interfaces let you type a dollar amount or percentage the strategy should use. That figure is rarely enforced by the exchange. If the key can see a larger balance or additional portfolios, the software itself must respect the smaller number. Once the software is compromised, that internal limit disappears. The only hard boundary is the one set on the exchange side through portfolio isolation or sub-account restrictions.
Practical Steps That Actually Reduce Exposure
After reviewing the recent case and earlier ones, a short list of habits stands out. None of them are exotic. Most exchanges already support them.
- Create a dedicated portfolio or sub-account that holds only the capital you are willing to risk with any automated tool. Keep the rest completely separate.
- Grant the absolute minimum permissions. Trade rights if needed, never transfer or withdrawal rights. Disable any funding or margin features the strategy does not require.
- Enable IP whitelisting wherever the exchange offers it and confirm the exact addresses the service will use. Update the list if the provider changes infrastructure.
- Set an expiry date on the key when the platform allows it. A credential that automatically dies after thirty or ninety days forces periodic review.
- Rotate keys after any security event or even on a regular schedule. Revoke the old one only after the new one is confirmed working.
- Monitor open orders and positions directly on the exchange interface rather than relying solely on the bot dashboard.
- Know whether the venue offers a dead-man switch or cancel-on-disconnect feature and enable it if available.
These steps do not make a strategy profitable. They simply limit the blast radius when something goes wrong. I’ve watched too many traders discover after the fact that a key they thought was tightly scoped still had access to far more capital than the bot interface displayed.
What Happens the Moment a Key Is Revoked
Revocation stops future requests. It does not rewrite history. Any fill that already occurred stays in the account. Any resting order that the exchange accepted remains active until it is canceled, expires, or fills. A derivatives position continues to gain or lose value according to market moves and margin rules. The bot’s previous stop-loss logic vanishes the instant the connection dies; only orders actually resting at the exchange continue to protect the position.
That is why both exchanges involved in the recent case told customers to log in and inspect everything themselves. A partially filled order can leave both a new holding and an outstanding remainder. A dashboard that lost connectivity may show outdated information. The exchange’s own order book and position screen become the only source of truth.
Reconnecting later introduces its own risk. A new key that suddenly sees open positions and resting orders created by the previous connection can behave unpredictably if the strategy was not written to handle that starting state. Cleaning the account first is almost always safer.
The Limits of Login Security Versus API Security
Two-factor authentication on the regular login page does not automatically protect an API connection. The key carries its own authentication method. Some exchanges offer additional two-factor options for API requests, but the client software must support them. Transfer permissions on certain platforms can even bypass interactive two-factor checks entirely. The two layers are independent and need separate attention.
QR codes and screenshots create another quiet leak. A photo that shows the private key or a QR code containing it can be as dangerous as the key itself. Covering a text field while leaving the code visible still hands an attacker everything needed to authenticate. Treating the entire credential set as sensitive material, including any visual representation, remains essential.
Reimbursement Does Not Reset Market Exposure
In the recent case the company moved quickly to repay the identified losses. That is welcome. It does not, however, reverse price moves that occurred after the unauthorized trades. A customer might receive compensation for the initial damage and still hold a position whose value continues to change. Closing or hedging that exposure remains the owner’s responsibility once direct control returns.
The same distinction applies to any future incident. Reimbursement addresses the unauthorized activity that the provider acknowledges. Ongoing market risk belongs to the account holder. Keeping the two concepts separate prevents unpleasant surprises when the dust settles.
Building Habits That Survive the Next Incident
No list of precautions eliminates risk. Automated trading will always carry the possibility of strategy failure or external compromise. The goal is to make sure a single bad key cannot reach the majority of an account’s capital or open uncontrolled leverage. Portfolio isolation, strict permission sets, IP restrictions, and regular key rotation form a practical baseline.
I also recommend a short monthly ritual: open the exchange API settings page, list every active key, confirm its permissions and last-used date, and revoke anything that is no longer needed. The exercise takes only a few minutes and surfaces forgotten credentials before they become a problem. When a third-party service experiences an incident, the same page becomes the first place to act rather than waiting for the provider to finish its own investigation.
Finally, treat any third-party dashboard as a convenience layer, never as the authoritative record. Balances, fills, and open risk live on the exchange. Checking them there after any interruption or security event remains the single most reliable habit a trader can adopt.
The recent episode did not invent a new category of risk. It simply made an existing one impossible to ignore. Trading permission is powerful. When that power sits inside software you do not fully control, the absence of withdrawal rights offers only partial comfort. Understanding the difference, and acting on it before the next incident, is the practical takeaway that remains after the reimbursements are complete and the services gradually return.
Markets will keep moving. New tools will keep appearing. The exchanges will continue offering granular permission controls. Using those controls with clear eyes is still the best defense available to anyone who chooses to automate part of their trading. The alternative is discovering too late that a locked withdrawal button never protected the value that quietly drained through ordinary trades.