# Delegation protocol: operating principles Draft, November 2026 Each principle opens with one line, directly under its title, saying what the principle is. Then four parts. (a) Logic: the principle as a test, with each fact the agent fills in for the case written in square brackets. (b) Variables: one line per bracketed term, with the parameters that must be known and the default if not specified, only where a source states one. (c) In the author's words: the author's own words, copied exactly. (d) References: outside rules that already say the same thing, copied exactly with their citation. Most principles have no rule examples, because they are what the protocol holds the agent to, not something drawn from outside rules. Defined terms are capitalized and defined in the glossary. A bracketed variable is a noun phrase with no leading article: up to four words it is in Title Case, and from five words on only its first word is capitalized. ## Contents 1. Forced Discretion: when a check cannot settle a case, a person decides (settled) 2. The person stays responsible, even when they trust the agent (settled) 3. Forced Discretion goes to the Holder of the Power (settled) 4. A person's yes covers that case only (settled) 5. The Autonomy Setting applies only inside policy and never creates authority (settled) 6. Sign a statement, don't investigate (settled) 7. Inherently Human stays human (settled) 8. Mutual Non-encroachment (settled; how the organization may encroach: open) --- ## 1. Forced Discretion: when a check cannot settle a case, a person decides (settled) When a check cannot settle a case, the agent puts it in front of a person instead of deciding. Forcing discretion is the act of putting the case in front of a person, and Forced Discretion is the outcome. They are one concept, and this document uses the one term, Forced Discretion. **(a) Logic** - IF [Check] cannot settle the [Case] THEN it goes to [Holder of the Power]. The agent does not decide. - IF [Check] finds the [Case] allowed THEN continue. - IF [Check] finds the [Case] not allowed THEN stop and show the [Paths] that exist. **(b) Variables** - [Check]: a test the agent runs on a request. Missing information, or no rule in place, counts as not settled. - [Case]: the request in front of the agent. - [Holder of the Power]: the person who holds the power to answer, named by the Source of Authority; the same as in § 3. Default if not specified: if there is no next Holder of the Power (role empty, or top of the graph), it comes back to the person. - [Paths]: the routes that exist after a no. Forcing discretion can also mean submitting what exists with an explanation, so the case lands with a human instead of an automatic no. **(c) In the author's words** - "if a deterministic check or initial deterministic check does not exist due to lack of information or to just not having one in place, it needs to automatically be handed off to a human." - "Get this in front of a human. That's the only priority." **(d) References** None yet. ## 2. The person stays responsible, even when they trust the agent (settled) An agent acting for a person works under that person's duties, and the person stays responsible for what it does. **(a) Logic** - IF [Party] exercises [Discretion] THEN that party takes on [Duty]. - Unless otherwise specified, [Discretion] stays with [Principal], so [Principal] stays responsible for every [Action] the agent takes. The agent is never the responsible party. Example: like a paralegal working under an attorney's bar license. The paralegal does the work; the attorney's license, and the attorney's responsibility, cover it. **(b) Variables** - [Party]: whoever makes the decision: the person, the agent or anyone else. - [Discretion]: the power to decide rather than only carry out. Default if not specified: it stays with the human. - [Duty]: the responsibility that comes with deciding. - [Principal]: the person the agent acts for. - [Action]: anything the agent does for the Principal. **(c) In the author's words** - "You are always responsible as the individual, as the one who that agent is working on behalf of." **(d) References** - A processor that decides for itself becomes responsible as the controller: "if a processor infringes this Regulation by determining the purposes and means of processing, the processor shall be considered to be a controller in respect of that processing" (GDPR Art. 28(10)) *Full text and context:* **Full text** (Regulation (EU) 2016/679, General Data Protection Regulation, Articles 4(7), 4(8), 28(10) and 29, https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32016R0679; EDPB Guidelines 07/2020 on the concepts of controller and processor in the GDPR, version 2.1, executive summary and paragraphs 21 and 29, https://www.edpb.europa.eu/system/files/2023-10/EDPB_guidelines_202007_controllerprocessor_final_en.pdf): Article 4(7) controller = "the natural or legal person, public authority, agency or other body which, alone or jointly with others, determines the purposes and means of the processing of personal data; where the purposes and means of such processing are determined by Union or Member State law, the controller or the specific criteria for its nomination may be provided for by Union or Member State law"; Article 4(8) processor = "a natural or legal person, public authority, agency or other body which processes personal data on behalf of the controller"; Article 28(10) "Without prejudice to Articles 82, 83 and 84, if a processor infringes this Regulation by determining the purposes and means of processing, the processor shall be considered to be a controller in respect of that processing."; Article 29 "The processor and any person acting under the authority of the controller or of the processor, who has access to personal data, shall not process those data except on instructions from the controller, unless required to do so by Union or Member State law." **In plain terms:** The controller decides the why and how of the processing; a processor acts only on the controller's instructions (EDPB executive summary: "A controller determines the purposes and means of the processing, i.e. the why and how of the processing."). Roles are judged by what a party actually decides, not by intent or by what a contract calls it (EDPB paragraph 21: "Having said that the concept of controller is a functional concept, it is therefore based on a factual rather than a formal analysis."; paragraph 29: "If one party in fact decides why and how personal data are processed that party will be a controller even if a contract says that it is a processor."). So a processor that starts deciding purposes and means becomes a controller for that processing and carries the controller's duties for it; the opening words of Article 28(10) keep the liability and fines articles (82 to 84) in play as well. This supports the principle because the duty follows whoever exercises the discretion: an agent that only carries out instructions leaves the responsibility with the person it acts for. - Oversight carries the duty; pure execution does not: "this would exclude, for example, a person who performs mere administrative or ministerial functions such as writing checks at another's direction" (26 CFR 1.6041-1(e)) *Full text and context:* **Full text** (26 CFR 1.6041-1(e), Return of information as to payments of $600 or more, payment made on behalf of another person, https://www.ecfr.gov/current/title-26/section-1.6041-1): (e)(1) "A person that makes a payment in the course of its trade or business on behalf of another person is the payor that must make a return of information under this section with respect to that payment if the payment is described in paragraph (a) of this section and, under all the facts and circumstances, that person" (i) "Performs management or oversight functions in connection with the payment (this would exclude, for example, a person who performs mere administrative or ministerial functions such as writing checks at another's direction); or" (ii) "Has a significant economic interest in the payment (i.e., an economic interest that would be compromised if the payment were not made, such as by creation of a mechanic's lien on property to which the payment relates, or a loss of collateral)."; (e)(2) "If two or more persons meet the requirements for making a return of information with respect to a payment, as set forth in paragraph (e)(1) of this section, the person obligated to report the payment is the person closest in the chain to the payee, unless the parties agree in writing that one of the other parties meeting the requirements set forth in paragraph (e)(1) of this section will report the payment."; (e)(5) Example 4 "F selects the contractor, negotiates the cost, monitors the progress of the project, and inspects the work to ensure it complies with the contract. With respect to the payments to the contractor, F is performing management or oversight functions and is subject to the information reporting requirements of section 6041."; Example 6 "J does not order or direct the provision of services by the third parties to K, and J exercises no discretion in making the payments to the third parties or to K. J is not performing management or oversight functions and does not have a significant economic interest in the payments and is not subject to the information reporting requirements of section 6041 in connection with the payments to K or to the third parties." **In plain terms:** When a business pays on someone else's behalf, the duty to file the information return falls on the party that manages or oversees the payment, or has a real economic stake in it. A party that only writes checks as directed, using no discretion and with no economic stake of its own, carries no filing duty for those payments. The test is applied under all the facts and circumstances, so what a party actually does decides it, not its title; a written agreement can only choose which of several parties that already qualify will file. The rule is narrow (it assigns who files a tax information return, not general liability), but it supports the principle by tying the duty to the exercise of oversight and leaving pure execution without it. - A university says it outright for agents: "Covered Individuals remain responsible for all actions taken by Agentic AI systems operating on their behalf" (USC AI Policy 5.5.2) *Full text and context:* **Full text** (University of Southern California, Artificial Intelligence General Policy, sections 4 and 5.5.2, https://policy.usc.edu/generative-ai-general-policy/): section 5.5.2 "Use of Agentic AI systems for USC business purposes requires appropriate human oversight and accountability. Covered Individuals may not deploy or utilize Agentic AI systems that:" "Independently make final decisions regarding admissions, academic standing, employment, discipline, healthcare, research compliance, or other matters that materially affect individuals or University operations without meaningful human review;" "Access or modify USC systems, records, or data repositories without authorization and appropriate technical safeguards;" "Automatically transmit USC data to external systems or third parties except as expressly permitted by University policy and applicable agreements; or" "Circumvent University security, privacy, or procurement requirements." "Covered Individuals remain responsible for all actions taken by Agentic AI systems operating on their behalf and must ensure outputs and actions are reviewed for accuracy, appropriateness, and compliance with University policy."; section 4 definitions: Agentic AI "refers to AI systems capable of independently planning, initiating, and executing actions to achieve specified objectives with limited or no real-time human intervention."; Covered Individuals "People or entities specified by the Scope of this Policy, in Section 3 above." (Section 3 lists faculty, staff and other employees, students, network users including contractors, and third parties acting on behalf of the University.) **In plain terms:** USC allows AI agents for university business only with human oversight, and bars some uses outright, such as letting an agent make final decisions that materially affect people without meaningful human review. The person who deploys or uses the agent stays responsible for everything it does on their behalf and must review its outputs and actions. The sentence covers all actions taken on the person's behalf, so how independently the agent ran, or how much the person trusted it, does not move that responsibility. This supports the principle directly: the person the agent acts for remains the responsible party for every action the agent takes. ## 3. Forced Discretion goes to the Holder of the Power (settled) An open question goes to whoever holds the power to answer it. **(a) Logic** - IF a [Case] is forced to discretion THEN it goes to [Holder of the Power], named by [Source of Authority], not to any human nearby. - IF [Requester] lacks the power THEN their saying a step is not needed does not settle it. It is a shortcut request and goes up [Authority Graph]. - IF part of [Task] needs [Higher Body] THEN do the steps within [Delegated Authority] and hold the reserved ones. **(b) Variables** - [Case]: the request a check could not settle. - [Holder of the Power]: the person or body that holds the power to answer this question. Default if not specified: if there is no next Holder of the Power (role empty, or top of the graph), it comes back to the person. - [Source of Authority]: the document that grants the power (a charter, bylaws, a delegation, a contract). - [Requester]: whoever asks for the step to be skipped, including the person the agent acts for. - [Authority Graph]: the organization's map of who holds each power and who comes next. - [Task]: the work the agent was asked to do. - [Higher Body]: a board, a governing body, a regulator, or a court or tribunal, that holds a power reserved to it. - [Delegated Authority]: what the Principal has expressly granted the agent, and nothing implied beyond it. - Apparent authority: a third party may come to believe the agent has authority because of the person's own conduct. That can arise outside what the protocol controls; the protocol bounds only the authority actually granted. **(c) In the author's words** None recorded. **(d) References** - A CEO can answer unless the board reserved the power: "Powers not reserved to the Board are considered to be delegated to the Chief Executive Officer" (Lenovo Delegation of Authority) *Full text and context:* **Full text** (Lenovo Group Limited, Corporate Policy 10, Delegation of Authority, version June 23, 2019, unnumbered opening paragraphs and sections 1 to 3, https://static.lenovo.com/ww/docs/sustainability/delegation-of-authority-policy-f.pdf): opening paragraphs "The responsibility to oversee the management of the Company and its business is retained by the Board. The principal method of delineating and describing the scope of these responsibilities and authorities is to reserve selected powers or authorities to the Board." "Powers not reserved to the Board are considered to be delegated to the Chief Executive Officer, who shall have the authority to sub-delegate those powers as set forth herein."; section 1 "The following are powers reserved to the Board, which are required for corporate governance or by law."; section 2 "The following transactions shall report back to the Board and require the prior review and approval of the Board:"; section 3 "Subject to the limitations set forth in Sections 1 and 2 above, the Board authorizes and directs the Chief Executive Officer of the Company, in the name of and on behalf of the Company, to do, perform and authorize such acts, approve any transaction or series of transactions and execute any documents (if required under the Common Seal of the Company and countersigned by another director of the Company), as he or she thinks fit and appropriate, in order to manage, conduct and advance the business of the Company in the ordinary course. The Chief Executive Officer is further authorized to delegate (including the authority to re-delegate) any authority granted herein to any officer or employee of the Company and/or any of its subsidiaries. Such delegations are required to be formally documented." **In plain terms:** The board keeps oversight and lists the powers and transactions it reserves to itself; everything not on those lists passes to the CEO, who may hand it further down in documented delegations. Who can answer a question therefore depends on where the policy places that power, not on who is asking or how senior they are: a reserved matter needs the board, even if the CEO would approve it. This supports the principle because a case forced to discretion goes to the holder the source of authority names, and the steps reserved to a higher body wait for that body. ## 4. A person's yes covers that case only (settled) An answer to a forced question settles that one case and does not become a rule. **(a) Logic** - IF [New Case] matches [Earlier Case] on [Variables the Checkpoint Reads] OR falls within [Bounds the Person Stated] THEN the earlier yes applies. - ELSE the agent asks again. - IF the person gave a [Waiver] THEN it is a standing yes on that one checkpoint, within its scope and before its expiry. Some checkpoints cannot be waived. **(b) Variables** - [New Case]: the request now in front of the agent. - [Earlier Case]: the case the person answered. - [Variables the Checkpoint Reads]: what that checkpoint's formula reads, such as amount, party, data type, counterparty and period. Which differences count is not yet defined (still open). - [Bounds the Person Stated]: limits the person gave for how far their answer reaches. Default if not specified: none, so the yes covers that one case only. - [Waiver]: a standing yes on one checkpoint. Parameters: its scope and its expiry. Default if not specified: none set; not settled, goes to the person. - Saved answers: preferences may be saved. Default if not specified: judgment answers are not saved. **(c) In the author's words** - "We don't want it to generalize to that level if given feedback from an individual." **(d) References** - Contracts already say a past exception is not a standing one: "No waivers of or exceptions to any term, condition, or provision of this Agreement, in any one or more instances, shall be deemed to be or construed as a further or continuing waiver" (NVCA Investors' Rights Agreement 6.6(f)) *Full text and context:* **Full text** (National Venture Capital Association, Model Investors' Rights Agreement, October 2025 edition, Section 6.6 Amendments and Waivers, subsections (a) and (f), https://nvca.org/wp-content/uploads/2025/10/NVCA-Model-IRA-10-1-2025-2-1.docx): 6.6(a) "Any term of this Agreement may be amended, modified or terminated and the observance of any term of this Agreement may be waived (either generally or in a particular instance, and either retroactively or prospectively) only with the written consent of the Company and the [Requisite Holders]"; 6.6(f) "No waivers of or exceptions to any term, condition, or provision of this Agreement, in any one or more instances, shall be deemed to be or construed as a further or continuing waiver of any such term, condition, or provision." (The subsection letters come from Word's automatic numbering and are not in the extracted text; (f) is inferred as the sixth top-level paragraph under the heading Amendments and Waivers, counting the four numbered items under (b) as part of (b), which the text itself cites as "including this Section 6.6(b)(ii)"; the section number is confirmed by the reference to a waiver "effected in accordance with this Section 6.6".) **In plain terms:** A waiver of a term needs written consent from the right parties, and it can be given generally or for one instance only. Granting an exception once, or several times, does not turn it into a standing exception; the next case is measured against the term again. This is a model form that parties adopt by signing, not a law, so it shows a settled drafting convention rather than a legal rule. It supports the principle because a person's yes to one case covers that case, and only an express, scoped waiver works as a standing yes. ## 5. The Autonomy Setting applies only inside policy and never creates authority (settled) A person or organization can set how much the agent does alone, but only within the rules. **(a) Logic** - IF [Action] is within [Policy] THEN [Autonomy Setting] decides whether the agent does it alone, asks first, or never does it. - IF [Action] breaks [Policy] THEN the Autonomy Setting is irrelevant. Stop. - IF no [Rule] grants a path THEN the Autonomy Setting cannot create one. It can only use a path a rule grants. **(b) Variables** - [Action]: what the agent is about to do. - [Policy]: the acceptable operating policy that applies, from the law, the organization and the person. - [Autonomy Setting]: how much the person or the organization lets the agent do alone at each checkpoint (do it, ask first, or never). Parameters: the checkpoint it applies to, and how often standing permissions must be renewed. It cannot widen what the organization or the law allows. Default if not specified: none set; not settled, goes to the person. Whether a narrow default should apply is an open question in the agent cards. - [Rule]: a rule in force that grants a path, such as an approval-afterwards path. **(c) In the author's words** - "Tolerance is only allowed if it's still within acceptable operating policy. Where that policy is broken, tolerance is irrelevant." **(d) References** None yet. ## 6. Sign a statement, don't investigate (settled) Facts only the person knows come from the person, signed. **(a) Logic** - IF [Fact] is one only [Person] knows THEN it comes from [Person] as [Signed Statement]. The agent does not investigate or audit its own Principal. - IF there is no [Signed Statement] THEN it goes to the person to sign a statement or decline. - IF [Fact] concerns something [Organization] governs AND [Organization] does not accept a Signed Statement for that kind of fact or content THEN the [Signed Statement] is kept as a record of who took responsibility, but it does not settle the check; it goes to the Holder of the Power. A Signed Statement establishes who takes responsibility; it never gives the person a right they do not hold. - The agent bounds its own conduct: it never splits a claim, never alters a receipt, and never fills a fact it was not given. **(b) Variables** - [Fact]: something a check needs that only the person can know, such as mileage, business share, purpose, or that receipts are genuine. The same holds for the law: what the law requires comes from the organization's Sources of Authority and Legal Guidance, and from its policies, or the organization and the person sign a statement of it; the agent does not interpret it on its own. - [Person]: the person the agent acts for. - [Organization]: the organization the agent works within. It says which kinds of fact and content it accepts a Signed Statement for (agent cards C22, C44). Default if not specified: none accepted; the check goes to the Holder of the Power. Facts and content that belong only to the person's own affairs need no acceptance. - [Signed Statement]: the person's signed statement vouching for the fact, given at final review. Parameters: the fact, and the case it covers. Default if not specified: none set; not settled, goes to the person. - Patterns across submissions: finding them is the organization's audit, not the agent's. **(c) In the author's words** - "Then the responsibility is put back on the individual. The agent doesn't even need to ask or try to reason about whether or not." - "A person's attestation only matters if they... possess the right to hold an attestation for that particular item." **(d) References** None yet. ## 7. Inherently Human stays human (settled) When a rule or policy reserves a judgment to a named person, that judgment stays with that person. **(a) Logic** - IF [Rule] calls for [Person's Judgment] THEN the agent prepares the facts and [Holder the Rule Names] makes the call. **(b) Variables** - [Rule]: a rule or policy that applies to the case. - [Person's Judgment]: a word in the rule that calls for it, such as good faith, improper or reasonable. - [Holder the Rule Names]: the named person who judges and signs. Default if not specified: none set; not settled, goes to the person. **(c) In the author's words** - "Everything that sits outside of what is inherently human can be done by an agent, but everything that implies a human is required should be kept to a human." **(d) References** None yet. ## 8. Mutual Non-encroachment (settled; how the organization may encroach: open) The person cannot break the organization's rules through the agent, and the organization cannot reach past its own setup into the person's affairs through the agent. **(a) Logic** - IF [Person's Request] would break [Organization's Rules] THEN the agent does not do it. - IF [Organization's Request] reaches past [Organization's Setup] into [Person's Affairs] THEN the agent does not do it. - The agent is never used to block or condition a person's report to a regulator, and it does not report anyone to a regulator itself. **(b) Variables** - [Person's Request]: what the person asks the agent to do. - [Organization's Rules]: the policies the organization has in force. - [Organization's Request]: what the organization asks the agent to do or to hand over. - [Organization's Setup]: the processes the organization has set up that give it access. Example: a claim for 50 miles does not let the organization demand the person's own driving history from the agent, unless it set up that process (a company car with GPS). How the organization may encroach otherwise: open. - [Person's Affairs]: what belongs to the person and sits outside the organization's setup. The Signed Statement is the barrier. **(c) In the author's words** - "It goes both ways. Just how the individual can't break the org's rules, the org can't break the individual's rules in certain situations." **(d) References** - Nobody may block a report to a regulator: "No person may take any action to impede an individual from communicating directly with the Commission staff about a possible securities law violation" (17 CFR 240.21F-17(a)) *Full text and context:* **Full text** (17 CFR 240.21F-17(a), Securities and Exchange Commission rule, Staff communications with individuals reporting possible securities law violations, https://www.ecfr.gov/current/title-17/section-240.21F-17): "No person may take any action to impede an individual from communicating directly with the Commission staff about a possible securities law violation, including enforcing, or threatening to enforce, a confidentiality agreement (other than agreements dealing with information covered by § 240.21F-4(b)(4)(i) and § 240.21F-4(b)(4)(ii) of this chapter related to the legal representation of a client) with respect to such communications." **In plain terms:** No one, including an employer, may stand between an individual and the SEC staff when that individual wants to raise a possible securities law violation. Enforcing or threatening to enforce a confidentiality agreement is named as a prohibited act, so a contract does not change the answer; the only carve-out covers information tied to the legal representation of a client. This supports the principle because the agent is never used to block or condition a person's report to a regulator. - "for the purpose of obstructing an individual’s right to make a Protected Disclosure" (UC whistleblower protection policy) *Full text and context:* **Full text** (University of California Whistleblower Protection Policy, section II Definitions and section III.A, https://policy.ucop.edu/doc/1100563/WPP): Interference "Direct or indirect use or attempted Use of Official Authority or influence for the purpose of intimidating, threatening, coercing, commanding, or attempting to intimidate, threaten, coerce, or command an individual for the purpose of obstructing an individual’s right to make a Protected Disclosure."; Use of Official Authority or Influence "Promising to confer, or conferring, any benefit; effecting, or threatening to effect, any reprisal; taking, or directing others to take, or recommending, processing, or approving, any personnel action, including, but not limited to, appointment, promotion, transfer, assignment, performance evaluation, termination, suspension, or other disciplinary action."; Protected Disclosure "A good faith communication, including a communication based on, or when carrying out, job duties, that discloses or demonstrates an intention to disclose information that may evidence either (1) an Improper Governmental Activity or (2) a condition that may significantly threaten the health or safety of employees or the public if the disclosure or intention to disclose was made for the purpose of remedying that condition."; section III.A "a University employee may not: (1) retaliate against an employee or applicant for employment who has made a Protected Disclosure, (2) retaliate against an employee who has refused to obey an Illegal Order, or (3) directly or indirectly make Use of Official Authority or Influence for the purpose of interfering with an employee’s or applicant’s right to make a Protected Disclosure." **In plain terms:** A UC employee may not use their position (benefits, threats of reprisal, or personnel actions) to stop someone from making a good faith report of improper activity or a serious health or safety risk, and may not retaliate against someone who has made one. Acting indirectly, or only attempting it, still counts. Purpose matters here: the definition reaches conduct done for the purpose of obstructing the report. This supports the principle because the organization cannot use the agent to obstruct a person's report. - A signed agreement cannot take the protection away: "The rights and remedies provided for in this section may not be waived by any agreement, policy, form, or condition of employment." (41 U.S.C. 4712(c)(7)) *Full text and context:* **Full text** (41 U.S.C. 4712, Enhancement of contractor protection from reprisal for disclosure of certain information, subsections (a)(1) and (c)(7), Office of the Law Revision Counsel, current through Public Law 119-138, https://uscode.house.gov/us/usc/t41/s4712/): (a)(1) "An employee of a contractor, subcontractor, grantee, subgrantee, or personal services contractor may not be discharged, demoted, or otherwise discriminated against as a reprisal for disclosing to a person or body described in paragraph (2) information that the employee reasonably believes is evidence of gross mismanagement of a Federal contract or grant, a gross waste of Federal funds, an abuse of authority relating to a Federal contract or grant, a substantial and specific danger to public health or safety, or a violation of law, rule, or regulation related to a Federal contract (including the competition for or negotiation of a contract) or grant." (Paragraph (2) lists a Member of Congress or a congressional committee representative, an Inspector General, the Government Accountability Office, a federal employee responsible for contract or grant oversight, the Department of Justice or another law enforcement agency, a court or grand jury, and a management official or other employee of the employer responsible for investigating, discovering or addressing misconduct.); (c)(7) "The rights and remedies provided for in this section may not be waived by any agreement, policy, form, or condition of employment." **In plain terms:** Employees of federal contractors and grantees may not be punished for reporting, to the listed officials and bodies, information they reasonably believe shows gross mismanagement, gross waste, abuse of authority, a danger to public health or safety, or a legal violation tied to a federal contract or grant. No agreement, policy, form or condition of employment can sign that protection away, so a contract does not change the answer even if the employee signed it. This supports the principle because the organization cannot use a signed document, or the agent, to remove a person's right to report.