Decentralised News · 2027 planning series · Reviewed 21 November 2026
What Matters
An alert about a DeFi exploit does not authorize an AI agent to move somebody else’s assets. A protocol can prepare an emergency-response arrangement, but the actual agreement, adoption record, covered assets, eligible actors and return conditions determine the question that needs review. A label such as “whitehat” cannot supply missing authority.
This article gives seven checks for an automated rescue proposal. It concerns defensive governance and evidence, not counter-hacking techniques. The local worksheet preserves missing and conflicting records without granting permission, interpreting a court’s likely decision or selecting a transaction. Its examples are invented and involve no live assets.
The Signal: a documented rescue framework is different from a kill switch
Security Alliance’s Safe Harbor overview describes a framework for protocols to pre-authorize qualifying interventions during active exploits. Its adoption guide separates scope definition, governance where required, registration and public terms. These are documentary steps that an emergency workflow can prepare in advance.
A kill switch addresses existing permissions and processes. A rescue arrangement addresses a different question: whether a particular responder may take a specified intervention under a particular set of terms. Stopping an agent, pausing an owned contract and intervening in another protocol should therefore have separate authority records.
DN Alpha Thesis
Prepare the authority record before the alert arrives. An emergency system is easier to review when it can identify an already approved scope, a dated agreement and an accountable response owner. The record should expose uncertainty instead of allowing speed, confidence or a claimed beneficial intention to substitute for permission.
1. Identify the adopted agreement and the organization behind it
Begin with the exact protocol or organization whose assets and operations are at issue. Preserve the governing agreement, its version and the evidence that the organization adopted it. A vendor’s marketing page about a framework establishes that the framework exists; it does not establish that every protocol using a similar technology adopted the same terms.
The official SEAL repository separates the legal agreement, a summary, frequently asked questions, release history and registry contracts. That structure makes version selection an explicit task. A current summary page and an agreement adopted earlier can refer to different material. Record both the authoritative document being relied on and the reason it applies.
For a DAO, distinguish the governance decision from the implemented registration and published scope. The self-adoption guide discusses these as separate steps. A passed proposal may describe an intention while another action still needs to establish the public operational record. The reviewer should retain evidence of what actually happened rather than assume that one completed stage implies every other stage.
Use references that another authorized reviewer can locate later. If a document changes, preserve the version assessed and the change history available to the team. A bare link to a page that can be edited is less informative than a dated source record. The worksheet accepts reference aliases; it does not fetch contracts, verify governance or authenticate an adoption.
Identify the responsible organizational role as well. An agent can retrieve documents, compare identifiers and highlight a missing term. Those tasks do not make the agent the organization’s legal decision-maker. The team should know who resolves a conflict when the governance record, registry entry and public explanation do not agree.
2. Distinguish a reportable vulnerability from the claimed emergency
A suspected flaw and an active exploit are different incident states. A framework designed for emergency intervention should not become a general permission to test live systems. Record the observation that triggered the alert, when it was observed and what conclusion the responsible reviewer actually drew.
SEAL’s whitehat guide ties its framework to specified active-exploit circumstances and excludes initiating the exploit oneself. It also discusses whether responsible disclosure is feasible. The article does not determine that any particular event meets those conditions. That determination needs the actual facts and applicable terms.
Immunefi’s ordinary platform rules separately prohibit whitehacking to save funds without the project’s express written consent in its dashboard. These rules demonstrate why a bounty listing should not be treated as a universal emergency authorization. A researcher must identify the operative route and any additional conditions, rather than blending separate programmes into a broader permission.
Preserve uncertainty in the alert. “The monitoring system detected an unusual withdrawal” is an observation. “An unauthorized exploit is underway” is an interpretation that requires support. A model-generated explanation can be part of the review, but it should not silently turn a noisy alert into an authority-triggering fact.
Prepare a documented escalation path for a suspected flaw that does not meet the emergency criteria. The useful design question is who receives the evidence and how its confidentiality is preserved. This article provides no operational method for exploiting, racing or diverting a live transaction.
3. Match the proposed response to the covered assets and effects
A protocol name is too broad to define the permitted intervention. Identify the chain, contract or account, relevant asset and scope rule that covers the proposed action. A project can have several deployments, factories, subsidiaries or upgraded components; a public brand does not establish that every associated object falls within one agreement.
The SEAL scope guidance documents asset scope, chain identifiers, recovery destinations, contacts and reward parameters. Its configurable structure supports a narrow conclusion: the actual adoption needs inspection. It does not justify borrowing another protocol’s scope or assuming that future deployments are included.
Describe the intended protective effect without expanding the action into a counterattack. The authority record should say which protected asset is involved and which approved response category is being considered. It should separately identify third-party systems and assets that are outside the reviewed boundary. A desirable outcome does not automatically authorize every technical path that might reach it.
Changes should reopen the review. A new deployment, upgraded contract, different chain or altered ownership can break the relationship between an earlier scope record and the proposed response. Preserve the relevant configuration date. An agent that finds a familiar project label should not reuse a stale scope assessment without checking the actual object.
DN’s governance-to-execution research explains the related problem of translating a mandate into the intended action. A rescue record needs the same discipline: what was approved, what action is proposed and what evidence connects the two. A syntactically valid instruction does not resolve a mismatch in meaning.
4. Establish the responder’s eligibility and remaining legal questions
Eligibility is a separate question from technical capability. A responder might be capable of taking an action while failing conditions in the relevant agreement. Record the actor or organizational role, the applicable eligibility review and any unresolved conflict. Use controlled records for identifying information; a public worksheet should contain aliases.
The whitehat guidance identifies independence from the original attacker and other conditions as material. It also states that the framework cannot protect against criminal charges outside its scope and that enforceability has uncertainties. DN therefore treats a safe-harbor label as a document to examine, never as a promise of universal immunity.
An organization using an automated responder should specify who operates it, who controls the relevant credentials and who can change its rules. Calling the software an autonomous agent does not answer those questions. A delegation record should identify the responsible party and the boundary within which the software assists.
Separate three review states: a requirement has been identified; a source of evidence has been recorded; and an authorized reviewer has assessed that evidence. The local instrument records declarations about those states but performs none of the assessment. A complete set of fields cannot cure an inaccurate declaration or an omitted legal condition.
Keep unresolved questions visible to the response owner. If the team cannot establish the adopted terms, the scope or the responder’s relationship to the incident, a favourable model confidence score does not resolve the problem. The next action should follow the organization’s approved escalation process, with the missing evidence identified clearly.
5. Separate asset recovery, custody and reward entitlement
A rescue proposal needs a documented destination for recovered assets and a controlled process for verifying that destination. Record the source of the recovery information and the responsible review. A message arriving during an incident should not silently replace the destination established by the governing records.
Custody also needs an explicit boundary. Who can control any assets during the permitted process, which actions are within the mandate and what evidence is retained? These are organizational questions to settle in advance. This article does not instruct a responder to move funds, select an address or retain a portion of assets.
Reward entitlement is another record. SEAL’s scope guidance describes configurable reward and retention terms, including situations in which recovered funds return before a reward is paid. The actual arrangement must be read. There is no universal percentage, cap, withholding right or payout date in DN’s worksheet.
Do not mix a claimed recovery amount with received compensation. A team’s internal incident account may record assets protected or returned, while a separate payment process determines a reward. If an eventual public report discusses a bounty, preserve its award and receipt evidence separately. DN’s AI-assisted bounty evidence method addresses that financial claim.
The distinction reduces a common reporting error: an intervention can have an asserted protective effect without establishing a paid reward, and a paid reward does not prove every technical claim in the incident narrative. Each conclusion should remain tied to the source that actually supports it.
6. Resolve notification and return terms from the operative record
Emergency conditions should not be reconstructed from remembered numbers. Record the applicable notification requirements, return conditions, timing triggers, permitted exceptions and evidence of completion. Identify which document supplies each term and which version was adopted for the protocol being reviewed.
The source review on 21 November 2026 found a useful warning against universal summaries. SEAL’s overview described a 72-hour return period, while its whitehat guide’s post-intervention section described six hours, or 48 hours with stated conditions. Those public summaries did not present the same timing. This article does not select either number as a universal rule.
The correct research task is to establish the governing agreement and adoption scope, then resolve the conflict through the responsible review process. Preserve the conflicting pages and dates as evidence of why clarification was needed. A software agent should not choose the longer period because it is more convenient, or the shorter one because a generic checklist says that is safer.
Timing has several distinct elements: the event that starts a period, the time at which that event occurred, the applicable unit or timezone and the evidence of the required action. A countdown built from the wrong trigger can look precise while answering the wrong question. For that reason, this worksheet records the timing source but contains no rescue-deadline calculator.
Notification evidence should distinguish an attempted message from an acknowledged contact where the applicable process requires that distinction. Return evidence should identify the actual observed state. Preserve unfinished obligations in the handover. An agent reporting “incident resolved” before those questions are documented can conceal work that still matters.
7. Keep independent oversight, revocation and the incident record
An automated response system should not be the sole author of its own authority, incident interpretation and final review. Assign a response owner and an independent way to suspend the system’s permitted actions. The architecture should make it possible to preserve evidence even when the automation’s recommendation is rejected.
DN’s kill-switch analysis distinguishes stopping a model from terminating its residual permissions, open orders and other executable paths. The related revocation research examines implementation-specific authority. An emergency rescue arrangement does not replace those controls; it adds another decision boundary that needs to be documented.
Record the sequence of observations, decisions, approvals and results with their sources. Keep the original alert separate from later interpretation. Preserve the agreement and configuration that existed at the time. A later revised policy should not be presented as though it governed an earlier intervention.
After the event, compare the intended response with what actually occurred. Which actions were proposed, which were permitted, which executed and which outcomes were observed? Identify remaining custody, communication, reward and disclosure questions. A retrospective should not infer success merely because the final balance looks better than a speculative worst case.
AI can help organize that evidence and identify inconsistencies. The claim that AI autonomously prevented a loss requires additional attribution and counterfactual evidence. DN’s incident-attribution ledger explains why a transaction trail and an actor or method claim must remain separate. This article establishes no empirical loss-prevention rate.
Worked case: clear technical alert, conflicting authority evidence
Imagine a fictional protocol, an internal responder and an unusual transaction alert. The team has an agreement alias, a version reference, a recorded adoption, a scope reference and a documented recovery destination. Its general monitoring procedure identifies a response owner. The source record still contains conflicting return summaries, and the team has not linked the applicable notification and return terms to the adopted version.
The worksheet reports the conflict and the missing references. It does not grant rescue permission because most fields are complete. It also does not assume that the alert establishes an active exploit. The immediate review question is concrete: which operative terms govern this proposed response, and what evidence establishes the relevant incident state?
| Evidence area | Synthetic record | Review consequence |
|---|---|---|
| Agreement and adoption | Aliases recorded | Underlying documents still need authentication and interpretation |
| Incident classification | Unusual activity reported | Emergency criteria remain to be assessed |
| Recovery destination | Reference recorded | No transaction is selected or authorized |
| Notification and return | Conflicting summaries | Resolve the adopted terms; no universal timer |
DN Rescue Authority Evidence Record
The record groups evidence into seven areas corresponding to the checks above. A group is documented only in the limited sense that the reader entered the required references. Counts describe missing documentation; they are not a rescue-readiness score. Unknown and conflicting declarations remain visible even when other references are complete.
DN Rescue Authority Evidence Record
Change the illustrative assumptions below. This worksheet runs locally in your browser. It records your declarations; it does not verify evidence, connect to accounts, submit instructions or grant approval.
The initial numbers are synthetic. Unknown evidence remains unresolved. Downloads contain the values you enter; use aliases and avoid private or identifying information.
Methodology, limitations and corrections
This is a primary-document review and a DN governance method, reviewed on 21 November 2026. DN did not assess a particular protocol’s legal adoption, investigate a live incident, validate a rescue transaction or determine a responder’s eligibility. The hypothetical record supplies no authority over a real asset.
The worksheet performs no network calls, contract interpretation, deadline calculation, target inspection or fund movement. It stores no underlying agreement and authenticates no identity. The source conflict is recorded as observed on the review date; later documentation may resolve it. Any update should identify the changed source and preserve the original evidence basis.
Use the method to prepare specific evidence requests and a reviewable handover. The final operational and legal decisions belong to the authorized parties applying the actual agreement and circumstances. A well-organized record can make their questions clearer; it cannot promise immunity, success, compensation or absence of harm.
Sources and evidence register
Primary sources reviewed on 21 November 2026. The article distinguishes documented facts, DN analysis and synthetic examples. Recheck changeable rules and configurations before acting.
- SEAL Safe Harbor overview — Framework scope and distinction from ordinary pre-exploit disclosure; summary timing is not treated as universal.
- SEAL Self-Adoption Guide — Scope, governance, registration and public adoption records.
- SEAL official agreement repository — Separates agreement, summaries, release history and registry implementation. No particular protocol adoption verified.
- SEAL Whitehat Rescue Guide — Eligibility and legal limitations; its timing summary differed from the overview during this review.
- SEAL Scope Terms — Protocol-specific scope, recovery, contact, identity and reward parameters.
- Immunefi platform rules — Ordinary bounty conduct and written-consent boundary for whitehacking to save funds.
Frequently asked questions
Does a bug-bounty programme automatically authorize an emergency rescue?
No. Identify the actual programme rules and any separate rescue agreement or express permission. A framework description or bounty listing alone does not establish authority for a particular intervention.
Does Safe Harbor provide universal legal immunity?
No universal immunity is established here. The reviewed whitehat guidance identifies conditions, enforceability uncertainties and legal matters outside the agreement’s scope. Actual facts, parties and applicable law still matter.
Which return deadline should an automated responder use?
Use the operative adopted agreement and applicable terms after resolving source conflicts. The reviewed public summaries differed, so this article and worksheet deliberately supply no universal deadline or countdown.
Can a complete worksheet authorize a rescue?
No. It records references and user declarations. It does not authenticate an adoption, interpret terms, establish an active exploit, assess eligibility or approve a transaction.
Is pausing my own system the same as rescuing another protocol’s funds?
They are different authority questions. Existing permissions and pause controls should have their own records; an intervention concerning another party’s assets needs its own applicable basis.
Can an AI agent decide whether the response qualifies?
It can help organize evidence and highlight conflicts, but this method does not delegate the operational or legal decision to the model. Identify the authorized response owner and applicable review process.




