출처 검증 리서치: CFTC SELF Custodial Wallet

Verified source facts

1. CFTC staff announced the no-action position on March 17, 2026. 2. The request involved Phantom Technologies and self-custodial wallet software facilitating access to registered futures entities. 3. Subject to conditions, staff would not recommend enforcement for failure to register as an introducing broker or associated person solely for the described activities. 4. The letter includes boundaries involving recommendations, order solicitation, fee relationships and trading privileges.

Comparison at a glance

1. Usage and measurement test: Document type

Stablecoin adoption cannot be inferred from a single gross counter. The document type dimension concerns staff no-action position. Define the economic job, unit of observation, time window and entities included before counting, then identify legal weight. Otherwise one routed transaction, exchange rebalance or contract hop may be mistaken for several independent users or payments.

Separate ledger events from economic outcomes. A transfer can be real on-chain yet still represent internal treasury movement, automated arbitrage, bridge routing, collateral management or redemption rather than a purchase. Official sources can frame the categories, but they do not resolve license confusion. The dataset and classification method therefore belong beside every headline number.

For document type, publish the numerator, denominator and exclusions that support the claim. Make the analyst identify legal weight, and compare at least one activity metric with one outcome metric relevant to staff no-action position. Because license confusion remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

2. Usage and measurement test: Covered entity

The covered entity dimension concerns specific requester. Define the economic job, unit of observation, time window and entities included before counting, then match legal person.

Official sources can frame the categories, but they do not resolve market-wide inference.

For covered entity, publish the numerator, denominator and exclusions that support the claim. Make the analyst match legal person, and compare at least one activity metric with one outcome metric relevant to specific requester. Because market-wide inference remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

3. Usage and measurement test: Custody

The custody dimension concerns self-custodial software. Define the economic job, unit of observation, time window and entities included before counting, then verify key control.

Official sources can frame the categories, but they do not resolve hidden custody.

For custody, publish the numerator, denominator and exclusions that support the claim. Make the analyst verify key control, and compare at least one activity metric with one outcome metric relevant to self-custodial software. Because hidden custody remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

4. Usage and measurement test: Recommendation

The recommendation dimension concerns neutral access or advice. Define the economic job, unit of observation, time window and entities included before counting, then audit prompts.

Official sources can frame the categories, but they do not resolve steering.

For recommendation, publish the numerator, denominator and exclusions that support the claim. Make the analyst audit prompts, and compare at least one activity metric with one outcome metric relevant to neutral access or advice. Because steering remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

5. Usage and measurement test: Order handling

The order handling dimension concerns software versus solicitation. Define the economic job, unit of observation, time window and entities included before counting, then trace flow.

Official sources can frame the categories, but they do not resolve intermediation.

For order handling, publish the numerator, denominator and exclusions that support the claim. Make the analyst trace flow, and compare at least one activity metric with one outcome metric relevant to software versus solicitation. Because intermediation remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

6. Usage and measurement test: Compensation

The compensation dimension concerns fee relationship. Define the economic job, unit of observation, time window and entities included before counting, then map incentives.

Official sources can frame the categories, but they do not resolve transaction-linked fee.

For compensation, publish the numerator, denominator and exclusions that support the claim. Make the analyst map incentives, and compare at least one activity metric with one outcome metric relevant to fee relationship. Because transaction-linked fee remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

7. Usage and measurement test: Counterparties

The counterparties dimension concerns registered FCM, IB or DCM. Define the economic job, unit of observation, time window and entities included before counting, then verify status.

Official sources can frame the categories, but they do not resolve unregistered route.

For counterparties, publish the numerator, denominator and exclusions that support the claim. Make the analyst verify status, and compare at least one activity metric with one outcome metric relevant to registered FCM, IB or DCM. Because unregistered route remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

8. Usage and measurement test: Privileges

The privileges dimension concerns market membership rights. Define the economic job, unit of observation, time window and entities included before counting, then check absence.

Official sources can frame the categories, but they do not resolve execution control.

For privileges, publish the numerator, denominator and exclusions that support the claim. Make the analyst check absence, and compare at least one activity metric with one outcome metric relevant to market membership rights. Because execution control remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

9. Usage and measurement test: Conditions

The conditions dimension concerns all stated requirements. Define the economic job, unit of observation, time window and entities included before counting, then maintain evidence.

Official sources can frame the categories, but they do not resolve model drift.

For conditions, publish the numerator, denominator and exclusions that support the claim. Make the analyst maintain evidence, and compare at least one activity metric with one outcome metric relevant to all stated requirements. Because model drift remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

10. Usage and measurement test: Stop rule

The stop rule dimension concerns business model differs. Define the economic job, unit of observation, time window and entities included before counting, then seek qualified review.

Official sources can frame the categories, but they do not resolve copycat reliance.

For stop rule, publish the numerator, denominator and exclusions that support the claim. Make the analyst seek qualified review, and compare at least one activity metric with one outcome metric relevant to business model differs. Because copycat reliance remains unobserved, the conclusion should describe measured behavior rather than declare universal adoption.

Decision scenarios

Scenario 1: when document type becomes decisive

Consider a reader whose main constraint is staff no-action position. The correct response is not to infer a universal platform or product winner. It is to identify legal weight, retain the evidence, and define a stop condition before funds or permissions are committed. If the evidence changes, the decision must be reopened. This produces a repeatable operating rule rather than a one-time opinion.

Scenario 2: when covered entity becomes decisive

Consider a reader whose main constraint is specific requester. It is to match legal person, retain the evidence, and define a stop condition before funds or permissions are committed.

Scenario 3: when custody becomes decisive

Consider a reader whose main constraint is self-custodial software. It is to verify key control, retain the evidence, and define a stop condition before funds or permissions are committed.

Scenario 4: when recommendation becomes decisive

Consider a reader whose main constraint is neutral access or advice. It is to audit prompts, retain the evidence, and define a stop condition before funds or permissions are committed.

Scenario 5: when order handling becomes decisive

Consider a reader whose main constraint is software versus solicitation. It is to trace flow, retain the evidence, and define a stop condition before funds or permissions are committed.

Scenario 6: when compensation becomes decisive

Consider a reader whose main constraint is fee relationship. It is to map incentives, retain the evidence, and define a stop condition before funds or permissions are committed.

Sources and editorial method

• Self-custodial wallet no-action position — CFTC
• Staff Letter 26-09 — CFTC
• Innovation at the CFTC — CFTC