Step 4: Full View

The Board's conclusion that AI-assisted drafting tools are not unethical per se must be qualified by a competence threshold that Engineer A did not meet with respect to the design documents. Code provisions I.2 and II.2.a require that engineers perform services only within areas of their competence, and this obligation extends to the tools they deploy. When an engineer uses a novel, unfamiliar AI drafting tool - one newly released to market with no prior experience on the engineer's part - and then conducts only a cursory, high-level review of its outputs before sealing and submitting engineering design documents, the engineer has not satisfied the competence standard that makes AI tool use ethically permissible in the first place. The Board's permissive conclusion about AI drafting tools implicitly assumes that the engineer possesses sufficient understanding of the tool's capabilities, limitations, and failure modes to exercise meaningful professional judgment over its outputs. Engineer A lacked that understanding entirely. The resulting design documents contained misaligned dimensions and omitted safety features required by local regulations - defects that a competent, engaged review would have identified. Accordingly, the ethical permissibility of AI-assisted drafting tools is conditional, not categorical: it depends on whether the engineer has sufficient competence with the tool and applies sufficient verification rigor to maintain genuine responsible charge over the work product.

The Board's finding that Engineer A's use of AI was partly unethical with respect to the design documents is further supported by the public safety dimension that the Board did not fully develop. Code provision I.1 places the safety, health, and welfare of the public as the paramount obligation of a licensed engineer, and this obligation is not merely aspirational - it is the foundational constraint against which all other professional judgments must be measured. The AI-generated design documents submitted by Engineer A contained omitted safety features required by local regulations. These omissions were not caught by Engineer A's cursory review and were only identified by Client W. Had Client W not conducted an independent technical review, those deficient documents could have proceeded to construction, creating a direct risk to public safety. The fact that the error was caught before construction does not retroactively satisfy the responsible charge standard; the standard requires that the engineer's own review be sufficient to ensure compliance, not that a client's independent review serve as the final safety check. Engineer A's sealing of documents containing regulatory safety omissions - after only a cursory review - therefore implicates not only Code provisions II.2.b and III.8.a regarding sealing and registration law compliance, but also the paramount public safety obligation of Code provision I.1. The ethical violation in the design phase is accordingly more serious than a mere procedural lapse in review thoroughness: it represents a failure of the core public protection function that professional licensure exists to serve.

The Board's conclusion that Engineer A has no universal ethical obligation to disclose AI use to Client W - analogizing AI tools to other engineering software - requires significant qualification in light of the specific facts of this case and must not be read as a blanket rule. The analogy to conventional engineering software breaks down in at least three respects. First, conventional design software such as CAD or finite element analysis tools operates deterministically on engineer-supplied inputs and produces outputs the engineer can fully audit; large language model AI generates probabilistic, non-deterministic text and design content whose provenance and accuracy the engineer cannot fully trace or verify. Second, the observable stylistic discontinuity in the report - which Client W independently detected, noting it read as if written by two different authors - created an implicit misrepresentation about the nature of the work product and its authorship. At the moment Client W raised that observation, Engineer A's silence became an act of omission that a reasonable client would regard as misleading, implicating Code provisions I.5 and III.3. Third, the design document defects - misaligned dimensions and omitted safety features - demonstrate that undisclosed AI-generated outputs in this case did reach a client and could have proceeded to construction without correction absent Client W's independent review. The Board's no-disclosure-obligation conclusion is therefore defensible only in circumstances where the engineer has exercised thorough, competent review of AI outputs and where no client inquiry or observable anomaly has created an affirmative duty to speak. In this case, neither condition was fully satisfied for the design documents, and the stylistic anomaly in the report created a specific moment at which silence was ethically problematic.

Beyond the Board's finding that Engineer A's use of AI in report writing was partly ethical and partly unethical, a critical and independent ethical breach exists that the Board did not explicitly address: Engineer A violated the client confidentiality obligation by uploading Client W's proprietary site data and groundwater monitoring information into an open-source AI platform without obtaining Client W's prior consent. Open-source AI platforms typically process and may retain user-submitted data in ways that expose it to third parties or incorporate it into training datasets, creating a foreseeable risk of disclosure beyond Engineer A's control. This breach of Code provision II.1.c stands entirely apart from questions about report quality, AI disclosure, or design document accuracy. A competent engineer deploying any third-party software tool - particularly a newly released, open-source platform with unknown data handling practices - bears an independent obligation to evaluate whether inputting confidential client data is permissible under the client relationship before acting. Engineer A's failure to seek Client W's consent before uploading that data constitutes a separate and self-standing ethical violation that the Board's analysis of report quality and AI transparency does not cure or subsume.

In response to Q101: Engineer A's upload of Client W's confidential site data and groundwater monitoring information into an open-source AI platform constitutes an independent and discrete ethical violation of Code provision II.1.c, entirely separate from any question about report quality or AI disclosure. The confidentiality obligation is not contingent on whether the resulting work product is accurate, polished, or ultimately beneficial to the client. By inputting proprietary client data into a publicly accessible AI system without obtaining Client W's prior consent, Engineer A exposed that information to potential third-party access, retention, or reuse by the AI platform - consequences Engineer A could not control or fully anticipate, particularly given their admitted unfamiliarity with the software. This breach stands on its own ethical foundation: the harm is the unauthorized exposure itself, not merely any downstream misuse. A competent engineer deploying a novel open-source tool with client data bears an affirmative obligation to investigate the data handling, storage, and privacy policies of that tool before use, and to obtain explicit client consent if any confidential information will be transmitted to a third-party system. Engineer A did neither. This violation is not remediated by the thoroughness of the subsequent report review, by the accuracy of the final work product, or by any disclosure or non-disclosure decision regarding AI authorship.

The Board's conclusion that AI-assisted drafting tools are not unethical per se must be qualified by a competence threshold that Engineer A did not meet with respect to the design documents. Code provisions I.2 and II.2.a require that engineers perform services only within areas of their competence, and this obligation extends to the tools they deploy. When an engineer uses a novel, unfamiliar AI drafting tool - one newly released to market with no prior experience on the engineer's part - and then conducts only a cursory, high-level review of its outputs before sealing and submitting engineering design documents, the engineer has not satisfied the competence standard that makes AI tool use ethically permissible in the first place. The Board's permissive conclusion about AI drafting tools implicitly assumes that the engineer possesses sufficient understanding of the tool's capabilities, limitations, and failure modes to exercise meaningful professional judgment over its outputs. Engineer A lacked that understanding entirely. The resulting design documents contained misaligned dimensions and omitted safety features required by local regulations - defects that a competent, engaged review would have identified. Accordingly, the ethical permissibility of AI-assisted drafting tools is conditional, not categorical: it depends on whether the engineer has sufficient competence with the tool and applies sufficient verification rigor to maintain genuine responsible charge over the work product.

The Board's analysis does not address a systemic professional vulnerability exposed by this case: Engineer A's over-reliance on AI tools was directly precipitated by the absence of the peer review and mentorship previously provided by Engineer B. When Engineer B retired, Engineer A lost not merely editorial guidance on technical writing but a substantive quality assurance mechanism that had been integral to Engineer A's professional practice. Rather than arranging an alternative peer review process - such as engaging a qualified colleague, a professional review service, or a subconsultant - Engineer A substituted an unfamiliar AI tool for that oversight function. This substitution was ethically inadequate for two independent reasons. First, AI tools are not peer reviewers: they do not apply independent professional judgment, cannot identify regulatory non-compliance from contextual knowledge, and cannot assume professional responsibility for the work. Second, the substitution required uploading confidential client data to an open-source platform, compounding the ethical problem. Code provision II.2.a's competence obligation and the broader duty of diligence implicit in responsible charge together suggest that when an engineer's established quality assurance mechanism becomes unavailable, the engineer bears an affirmative obligation to arrange a functionally equivalent alternative before undertaking complex, high-stakes engagements - not to proceed with an untested technological substitute. The NSPE Code of Ethics does not currently provide explicit guidance on peer review succession planning, and this case illustrates that such guidance would meaningfully serve the profession.

The Board's finding that Engineer A's use of AI was partly unethical with respect to the design documents is further supported by the public safety dimension that the Board did not fully develop. Code provision I.1 places the safety, health, and welfare of the public as the paramount obligation of a licensed engineer, and this obligation is not merely aspirational - it is the foundational constraint against which all other professional judgments must be measured. The AI-generated design documents submitted by Engineer A contained omitted safety features required by local regulations. These omissions were not caught by Engineer A's cursory review and were only identified by Client W. Had Client W not conducted an independent technical review, those deficient documents could have proceeded to construction, creating a direct risk to public safety. The fact that the error was caught before construction does not retroactively satisfy the responsible charge standard; the standard requires that the engineer's own review be sufficient to ensure compliance, not that a client's independent review serve as the final safety check. Engineer A's sealing of documents containing regulatory safety omissions - after only a cursory review - therefore implicates not only Code provisions II.2.b and III.8.a regarding sealing and registration law compliance, but also the paramount public safety obligation of Code provision I.1. The ethical violation in the design phase is accordingly more serious than a mere procedural lapse in review thoroughness: it represents a failure of the core public protection function that professional licensure exists to serve.

In response to Q102: Engineer B's retirement did not merely create an inconvenience for Engineer A - it removed the primary quality assurance mechanism upon which Engineer A had structurally depended for professional-grade output, particularly in technical writing. When that mechanism was removed, Engineer A faced a dual-scope engagement of meaningful complexity: a comprehensive contaminant characterization report requiring synthesis of groundwater monitoring data, and engineering design documents for infrastructure modifications. Rather than arranging an alternative peer review process - such as engaging a qualified colleague, contracting a third-party reviewer, or consulting with a professional organization - Engineer A substituted an unfamiliar, newly released open-source AI tool for that professional oversight. This substitution was not ethically neutral. The NSPE Code's competence provisions (I.2 and II.2.a) require engineers to undertake assignments only when qualified, and qualification encompasses not only technical domain knowledge but also the professional infrastructure necessary to deliver work of adequate quality. An engineer who knows they have a recognized weakness in a critical deliverable component, who has lost their primary quality assurance resource, and who then deploys an untested tool as a replacement - without any independent verification of that tool's reliability - has not satisfied the competence standard. Engineer A had an independent ethical obligation to arrange alternative peer review before proceeding, and the failure to do so compounded every subsequent deficiency in both the report and the design documents.

The Board's conclusion that Engineer A has no universal ethical obligation to disclose AI use to Client W - analogizing AI tools to other engineering software - requires significant qualification in light of the specific facts of this case and must not be read as a blanket rule. The analogy to conventional engineering software breaks down in at least three respects. First, conventional design software such as CAD or finite element analysis tools operates deterministically on engineer-supplied inputs and produces outputs the engineer can fully audit; large language model AI generates probabilistic, non-deterministic text and design content whose provenance and accuracy the engineer cannot fully trace or verify. Second, the observable stylistic discontinuity in the report - which Client W independently detected, noting it read as if written by two different authors - created an implicit misrepresentation about the nature of the work product and its authorship. At the moment Client W raised that observation, Engineer A's silence became an act of omission that a reasonable client would regard as misleading, implicating Code provisions I.5 and III.3. Third, the design document defects - misaligned dimensions and omitted safety features - demonstrate that undisclosed AI-generated outputs in this case did reach a client and could have proceeded to construction without correction absent Client W's independent review. The Board's no-disclosure-obligation conclusion is therefore defensible only in circumstances where the engineer has exercised thorough, competent review of AI outputs and where no client inquiry or observable anomaly has created an affirmative duty to speak. In this case, neither condition was fully satisfied for the design documents, and the stylistic anomaly in the report created a specific moment at which silence was ethically problematic.

In response to Q103: When Client W directly observed that the report appeared to have been written by two different authors - a stylistically inconsistent observation that was, in fact, an accurate description of the report's dual-origin nature - Engineer A's silence in that moment was not ethically neutral. Code provisions I.5 and III.3 prohibit deceptive acts and conduct that deceives the public or clients. Deception does not require an affirmative false statement; it can arise from deliberate silence in circumstances where a reasonable client would expect disclosure and where the omission creates or sustains a materially false impression. Client W's comment was a direct, specific observation that implicitly invited clarification about the report's authorship. A client who is told their report reads as if written by two people is, in practical terms, asking why. Engineer A's failure to respond honestly - by acknowledging that AI software had generated the more polished sections - allowed Client W to proceed under the false impression that the entire report was the product of Engineer A's own professional authorship. This silence, in context, constitutes a deceptive act under I.5 and conduct that deceives under III.3, independent of whether disclosure was required before submission. The moment of Client W's observation created a discrete, time-specific obligation to clarify, and Engineer A's failure to do so transformed a prior omission into an active, ongoing misrepresentation.

The Board's analysis does not address a systemic professional vulnerability exposed by this case: Engineer A's over-reliance on AI tools was directly precipitated by the absence of the peer review and mentorship previously provided by Engineer B. When Engineer B retired, Engineer A lost not merely editorial guidance on technical writing but a substantive quality assurance mechanism that had been integral to Engineer A's professional practice. Rather than arranging an alternative peer review process - such as engaging a qualified colleague, a professional review service, or a subconsultant - Engineer A substituted an unfamiliar AI tool for that oversight function. This substitution was ethically inadequate for two independent reasons. First, AI tools are not peer reviewers: they do not apply independent professional judgment, cannot identify regulatory non-compliance from contextual knowledge, and cannot assume professional responsibility for the work. Second, the substitution required uploading confidential client data to an open-source platform, compounding the ethical problem. Code provision II.2.a's competence obligation and the broader duty of diligence implicit in responsible charge together suggest that when an engineer's established quality assurance mechanism becomes unavailable, the engineer bears an affirmative obligation to arrange a functionally equivalent alternative before undertaking complex, high-stakes engagements - not to proceed with an untested technological substitute. The NSPE Code of Ethics does not currently provide explicit guidance on peer review succession planning, and this case illustrates that such guidance would meaningfully serve the profession.

Engineer A's failure to cite the professional journal articles used to cross-check AI-generated content, and the absence of any attribution for the AI-generated text itself, raises an underexamined concern about the evidentiary integrity of a technical report that may inform regulatory decisions or remediation actions. Code provision III.9 requires engineers to give credit for engineering work to those to whom credit is due. While this provision is most commonly applied to prevent engineers from claiming credit for others' work, it also carries an affirmative dimension: a technical report submitted in a professional capacity implicitly represents that its intellectual content reflects the engineer's own analysis and judgment. Where substantial portions of the report's prose and synthesis were generated by an AI system, and where the factual cross-checking relied on professional journal articles that are not cited, the report's evidentiary foundation is obscured. Regulators, future engineers, or legal proceedings relying on the report cannot assess the quality of the underlying analysis, trace its sources, or evaluate the reliability of the AI-generated synthesis. This is particularly consequential for a report addressing an emerging contaminant of concern, where the scientific basis for conclusions may be contested and where the report may serve as a foundational document for remediation planning or regulatory compliance. The absence of attribution and citation therefore undermines not only intellectual honesty in authorship but also the professional reliability and traceability of the work product itself.

In response to Q104: Engineer A's failure to cite the professional journal articles used to cross-check AI-generated content raises a concern under Code provision III.9, which requires engineers to give credit for engineering work to those to whom credit is due. While III.9 is most commonly applied to crediting the work of other engineers, its underlying principle - that the intellectual and evidentiary foundations of professional work must be honestly attributed - extends to the sources that substantiate technical conclusions. In a report that may inform regulatory decisions or remediation actions affecting public health and environmental safety, the absence of citations to the scientific literature used to verify AI-generated claims is not merely a stylistic deficiency. It deprives Client W, regulators, and any subsequent reviewers of the ability to independently assess the evidentiary basis for the report's conclusions, to identify the scope and currency of the literature consulted, and to evaluate whether the cross-checking process was adequate. This omission undermines the epistemic integrity of the report as a professional document. Furthermore, in the context of an emerging contaminant of concern - a category of substance where scientific understanding is actively evolving - the failure to ground conclusions in cited, verifiable sources creates a foreseeable risk that outdated, incomplete, or AI-hallucinated information could go undetected by downstream users who rely on the report's apparent professional authority.

In response to Q201: A genuine tension exists between the principle that professional competence in report writing can be satisfied through thorough post-generation verification and the principle of intellectual honesty in authorship. The Board concluded that Engineer A's thorough review of the AI-generated report text was sufficient to render that use of AI ethical. However, this conclusion does not fully resolve the authorship integrity question. When an engineer applies their professional seal to a document, they represent to the client and to the public that the work reflects their professional judgment, expertise, and authorship. The seal is not merely a quality certification - it is an assertion of intellectual ownership and responsible charge. A report whose prose was substantially composed by a non-human language model, and whose authorship was personalized only through minor wording adjustments, does not straightforwardly satisfy that representation, even if every factual claim has been verified. The verification process confirms accuracy; it does not transform AI-generated prose into the engineer's own professional expression. These two principles can be reconciled only if the engineering profession explicitly adopts a framework - which it has not yet done - that defines AI-assisted authorship as a recognized and disclosed mode of professional work product creation. Absent such a framework, the tension remains unresolved, and the Board's conclusion on report ethics should be understood as provisional rather than definitive.

The tension between Professional Competence Satisfied for Report Writing and Intellectual Honesty in Authorship was left substantively unresolved by the Board. The Board accepted that Engineer A's thorough factual verification of AI-generated text satisfied the competence dimension of responsible charge for the report, but it did not squarely confront the authorship dimension: when an engineer personalizes AI-generated prose with only minor wording adjustments and submits it under a professional seal without attribution, the seal implicitly represents that the engineer is the intellectual author of the work product. These two principles pull in opposite directions - competence review can be satisfied by rigorous fact-checking, but intellectual honesty in authorship requires that the origin of the substantive prose be accurately represented. The case teaches that competence and authorship are distinct professional obligations, and that satisfying one does not discharge the other. A fully ethical resolution would have required Engineer A to either disclose the AI's generative role or to rewrite the report in their own voice after verification, rather than treating minor wording edits as sufficient to claim authorship.

In response to Q202: The tension between Responsible Charge Engagement and Competence Assurance Under Novel Tool Adoption is not merely theoretical - it is demonstrated concretely by the outcome in this case. Engineer A applied their professional seal to AI-generated design documents after only a cursory, high-level review. The professional seal carries a legal and ethical certification that the engineer has exercised responsible charge over the work: that they understand its content, have directed its preparation, and can stand behind its technical adequacy. A cursory review of output generated by a novel AI drafting tool - one with which Engineer A had no prior experience and whose generative logic Engineer A did not fully understand - cannot satisfy that standard. The subsequent discovery of misaligned dimensions and omitted safety features required by local regulations confirms that the cursory review was substantively inadequate. Code provision II.2.b prohibits engineers from affixing their signatures to plans dealing with subject matter in which they lack competence. Competence here encompasses not only domain knowledge in groundwater infrastructure design, but also sufficient understanding of the AI tool's outputs to certify their reliability. Engineer A possessed the former but demonstrably lacked the latter. The seal, in this context, was affixed in violation of II.2.b, and the tension between these two principles is resolved against Engineer A: responsible charge cannot be satisfied by reviewing outputs from a tool whose behavior the reviewing engineer does not adequately understand.

The tension between Responsible Charge Engagement and Competence Assurance Under Novel Tool Adoption was resolved against Engineer A in the design document context, but the resolution reveals a deeper principle hierarchy: when an engineer applies a professional seal, the seal does not merely certify that the engineer reviewed the output - it certifies that the engineer exercised personal, informed judgment over the generative process itself. Because Engineer A had no prior experience with the AI drafting tool and did not understand its full functionality, a cursory high-level review was structurally incapable of satisfying responsible charge, regardless of how much time was spent. The case teaches that the standard of review required to satisfy responsible charge scales inversely with the engineer's familiarity with the generative tool: the less the engineer understands how the tool produces its output, the more rigorous the independent verification must be. Deploying an unfamiliar AI tool is not ethically equivalent to deploying familiar software; it introduces an epistemic gap that only deeper review - not a high-level scan - can close. Public Welfare Paramount ultimately overrides both efficiency and tool novelty as a justification for reduced oversight, particularly where safety-critical omissions in design documents could reach construction.

The tension between Client Data Confidentiality in AI Tool Use and Mentorship Continuity and Succession Planning exposes a systemic vulnerability that the Board's conclusions do not address: Engineer A's loss of Engineer B's peer review created professional pressure to substitute AI assistance for human oversight, but the only available AI tool was open-source, meaning that satisfying the need for quality assurance necessarily required exposing Client W's confidential site data and groundwater monitoring information to a public platform without prior consent. This creates a structural conflict in which the engineer cannot simultaneously honor the confidentiality obligation and use the available compensating mechanism. The case teaches that this conflict is not resolvable by choosing one principle over the other after the fact - it is resolvable only by proactive planning before the engagement begins. The principle of Mentorship Continuity and Succession Planning, read alongside the confidentiality obligation under Code provision II.1.c, implies that when a primary quality assurance mechanism is lost, the engineer's first obligation is to identify a compliant replacement - whether a qualified peer reviewer, a privacy-compliant AI platform, or a scope limitation - before accepting work that cannot be competently and confidentially performed alone. Engineer A's failure to engage in that prior planning rendered the confidentiality breach not merely a procedural lapse but a foreseeable consequence of an inadequately structured professional practice.

In response to Q204: The Board's conclusion that there is no universal ethical obligation to disclose AI use is placed under significant strain by the facts of this case. The principle that public welfare is paramount - Code provision I.1 - is not merely aspirational; it functions as a constraint on every other professional decision an engineer makes. In this case, AI-generated design documents containing omitted safety features required by local regulations were submitted to Client W under Engineer A's professional seal. Had Client W not independently identified these deficiencies, the documents could have proceeded toward construction in a non-compliant and potentially dangerous state. The Board's general conclusion about disclosure is grounded in an analogy to other software tools used in engineering practice - an analogy that may hold when the tool is well-understood, widely validated, and used within established professional norms. It does not hold with equal force when the tool is newly released, unfamiliar to the practitioner, and demonstrably capable of generating safety-critical omissions that a cursory review failed to catch. In such circumstances, the public welfare principle does not merely permit disclosure - it may affirmatively require it, because disclosure enables the client and downstream reviewers to apply appropriate scrutiny to outputs whose reliability has not been professionally validated. The Board's conclusion on disclosure should therefore be understood as conditional: it applies when AI tools are used competently and their outputs are rigorously verified, not when they are deployed as substitutes for professional judgment with only superficial review.

The tension between Responsible Charge Engagement and Competence Assurance Under Novel Tool Adoption was resolved against Engineer A in the design document context, but the resolution reveals a deeper principle hierarchy: when an engineer applies a professional seal, the seal does not merely certify that the engineer reviewed the output - it certifies that the engineer exercised personal, informed judgment over the generative process itself. Because Engineer A had no prior experience with the AI drafting tool and did not understand its full functionality, a cursory high-level review was structurally incapable of satisfying responsible charge, regardless of how much time was spent. The case teaches that the standard of review required to satisfy responsible charge scales inversely with the engineer's familiarity with the generative tool: the less the engineer understands how the tool produces its output, the more rigorous the independent verification must be. Deploying an unfamiliar AI tool is not ethically equivalent to deploying familiar software; it introduces an epistemic gap that only deeper review - not a high-level scan - can close. Public Welfare Paramount ultimately overrides both efficiency and tool novelty as a justification for reduced oversight, particularly where safety-critical omissions in design documents could reach construction.

In response to Q301: From a deontological perspective, Engineer A did not fulfill their duty of candor toward Client W. Kantian deontological ethics evaluates the moral worth of an action by reference to the maxim underlying it and whether that maxim could be universalized without contradiction. The maxim implicit in Engineer A's conduct - that an engineer may submit AI-generated work products under their professional seal without disclosing the AI's role, provided the outputs are verified for accuracy - cannot be universalized without undermining the foundational trust relationship between licensed professionals and their clients. If all engineers adopted this maxim, the professional seal would cease to function as a reliable signal of personal authorship and responsible charge, and clients would be systematically deprived of information material to their assessment of the work product's provenance and reliability. Furthermore, the duty of candor is not contingent on outcome: it is not satisfied by the fact that the report was accurate or that the design errors were caught. Deontological ethics holds that the duty to be honest with those who rely on one's professional representations exists independently of whether the deception caused harm. Engineer A's silence about AI's role - particularly in the face of Client W's direct observation about the report's stylistic inconsistency - constitutes a breach of the duty of candor that is not remediated by the quality of the final work product.

In response to Q302: From a deontological perspective, Engineer A breached their categorical duty to maintain Responsible Charge by sealing engineering design documents that contained safety omissions and dimensional errors they had only cursorily reviewed. Responsible Charge is not a procedural formality - it is a substantive professional and ethical duty that requires the engineer to have directed the work, to understand its content, and to be able to certify its technical adequacy. The professional seal is the outward expression of that duty, and affixing it to documents that have not been adequately reviewed is a categorical violation regardless of intent or outcome. From a deontological standpoint, the duty is breached at the moment of sealing, not at the moment of harm. The fact that Client W identified the errors before construction does not retroactively satisfy the Responsible Charge obligation; it merely prevented the consequences from being worse. Code provision II.2.b makes this categorical character explicit: engineers shall not affix their signatures to plans dealing with subject matter in which they lack competence. Engineer A's unfamiliarity with the AI drafting tool's outputs, combined with a cursory review that failed to detect regulatory non-compliance, establishes that the competence threshold was not met at the time of sealing. The deontological analysis therefore yields a clear conclusion: the duty was breached, independently of any consequentialist assessment of harm.

In response to Q303: From a consequentialist perspective, the harm produced by Engineer A's cursory review of AI-generated design documents - resulting in misaligned dimensions and omitted safety features required by local regulations - does outweigh the efficiency benefits gained from using AI-assisted drafting tools in the design phase, and this outcome is ethically significant even though the errors were caught before construction. Consequentialist analysis evaluates actions by their expected outcomes, including foreseeable risks. A competent engineer deploying a novel, untested AI drafting tool for safety-critical infrastructure design, with no prior experience and only a cursory review process, creates a foreseeable probability of undetected errors reaching construction. The actual outcome - regulatory non-compliance and safety omissions - was not an improbable accident; it was a predictable consequence of an inadequate verification process applied to an unreliable generative tool. The efficiency gain from AI-assisted drafting is real but modest relative to the risk: the time saved in initial document generation was offset by the need for revision, the erosion of client trust, and the potential - had Client W not been diligent - for construction of non-compliant infrastructure. Consequentialist ethics does not require that harm actually occur to render a decision unethical; it requires that the expected value of the action, accounting for foreseeable risks, be negative. Here, the expected value of deploying an unfamiliar AI tool with cursory review for safety-critical design work was negative at the time of the decision, and the actual outcome confirms that assessment.

In response to Q305: From a virtue ethics perspective, Engineer A did not exhibit the prudence and professional humility expected of a competent engineer in choosing to deploy a novel, unfamiliar AI drafting tool as a substitute for the mentorship and peer review previously provided by Engineer B. Prudence - the virtue of practical wisdom applied to professional decision-making - requires an engineer to accurately assess their own capabilities and limitations, to recognize the boundaries of their competence, and to seek appropriate resources when those boundaries are approached. Engineer A's self-acknowledged weakness in technical writing, combined with the loss of Engineer B's quality assurance function, created a situation that called for heightened caution and deliberate compensatory measures. Instead, Engineer A responded by introducing a second source of uncertainty: an AI tool that was new to the market, open-source, and entirely unfamiliar to Engineer A. Professional humility would have led Engineer A to recognize that substituting one unknown - AI-generated output - for a known quality assurance resource - Engineer B's expert review - does not reduce professional risk; it compounds it. A prudent engineer in Engineer A's position would have sought an alternative qualified peer reviewer, disclosed the limitation to Client W, or scoped the engagement to match their verified capabilities. The choice to proceed without these safeguards reflects not merely a procedural lapse but a deficit in the practical wisdom that the engineering profession requires of its licensed practitioners.

In response to Q306: From a consequentialist perspective, Engineer A's decision to input Client W's confidential site data into open-source AI software without prior consent created a foreseeable risk of harm to Client W's proprietary interests that outweighs the drafting efficiency gained, and that risk calculus should have been apparent to a competent engineer before acting. Open-source AI platforms are, by their nature, systems whose data handling, retention, training data incorporation, and third-party access policies are not under the control of the user. A competent engineer - particularly one engaged in environmental consulting involving site-specific groundwater data that may have regulatory, litigation, or competitive sensitivity - bears a professional obligation to investigate how any third-party system will handle client data before transmitting it. The efficiency benefit of AI-assisted drafting is real but bounded: it accelerates initial document generation. The risk created by uploading confidential client data to an unvetted public platform is potentially unbounded: it includes regulatory exposure, competitive harm, litigation risk, and reputational damage to Client W. A consequentialist analysis that assigns even a modest probability to these harms - and a competent engineer should have assigned a non-trivial probability - yields a negative expected value for the decision to use open-source AI without consent. The fact that Engineer A was unfamiliar with the AI software's full functionality, including its data handling practices, does not mitigate this conclusion; it reinforces it, because proceeding under conditions of ignorance about foreseeable risks is itself a consequentialist failure.

In response to Q304: From a virtue ethics perspective, Engineer A did not demonstrate the professional integrity and intellectual honesty expected of a licensed engineer in the report authorship process. Virtue ethics evaluates conduct by reference to the character traits and dispositions that a person of practical wisdom - a phronimos - would exhibit in the relevant professional role. A licensed engineer of good character, confronted with a recognized weakness in technical writing and the loss of their primary quality assurance resource, would seek transparent solutions: engaging a peer reviewer, disclosing limitations to the client, or explicitly attributing AI assistance in the work product. Engineer A instead chose a path that preserved the appearance of unassisted professional authorship while relying substantially on AI-generated prose. The minor wording adjustments made to personalize the content do not constitute the kind of intellectual engagement that transforms another's expression into one's own. A person of practical wisdom would recognize that submitting AI-generated text under a professional seal - without attribution, and in the face of a client's direct observation about stylistic inconsistency - is not merely a procedural omission but a failure of intellectual honesty. The virtue of integrity requires consistency between one's professional representations and the actual nature of one's work. Engineer A's conduct fell short of that standard, regardless of the report's factual accuracy.

In response to Q402: If Engineer A had conducted a rigorous, line-by-line technical review of the AI-generated design documents - equivalent in thoroughness to the review applied to the report - the safety omissions and dimensional errors would very likely have been identified before submission to Client W, and such a review would have been substantially more likely to satisfy the Responsible Charge standard. The case establishes a clear asymmetry: Engineer A's thorough review of the report was sufficient to catch factual inaccuracies and verify content quality, while the cursory review of the design documents was not sufficient to detect regulatory non-compliance and dimensional errors. This asymmetry suggests that the review process, not the use of AI per se, was the determinative variable in the design document failure. A rigorous review - one that checked each dimension against site survey data, verified each specification against local regulatory requirements, and confirmed the presence of all required safety features - would have functioned as an adequate Responsible Charge mechanism even for AI-generated outputs, provided the reviewing engineer possessed the domain competence to evaluate what they were reviewing. Engineer A possessed that domain competence in groundwater infrastructure design. The ethical failure was therefore not the use of AI drafting tools, but the decision to apply a cursory rather than rigorous review standard to safety-critical outputs from an untested tool. This counterfactual reinforces the Board's conclusion that AI-assisted drafting is not unethical per se, while clarifying that the adequacy of the review process is the critical ethical variable.

In response to Q404: If Engineer A had explicitly cited the use of AI software in the report - identifying which sections were AI-generated and which were independently authored - Client W's observation that the report read as if written by two different authors would have been resolved rather than raised as a concern. The stylistic inconsistency that Client W detected was, in fact, an accurate artifact of the report's dual-origin nature: AI-generated prose tends to exhibit a characteristic uniformity and polish that differs from the more variable style of human technical writing, particularly from an engineer who self-identifies as less confident in technical writing. Explicit attribution would have provided Client W with a framework for understanding and contextualizing that observation, transforming a source of unease into a transparent feature of the work product. However, explicit attribution would also have raised a different set of questions: it would have invited Client W to scrutinize the AI-generated sections more carefully, to inquire about the AI tool's data handling practices, and potentially to raise concerns about the confidential data exposure that had already occurred. In this sense, disclosure would have been simultaneously clarifying and consequential - it would have resolved the authorship ambiguity while potentially surfacing the deeper confidentiality violation. This counterfactual suggests that the ethical case for disclosure is stronger than the Board's agnostic conclusion implies: transparency about AI use not only serves intellectual honesty but also enables clients to exercise informed oversight of work products that may have been generated under conditions they would not have approved.

In response to Q401: If Engineer A had disclosed their intended use of open-source AI software to Client W before beginning work, and Client W had withheld consent to upload confidential site data to a public AI platform, Engineer A would have faced a clear ethical fork: either decline the use of open-source AI tools entirely, or identify a privacy-compliant alternative - such as an enterprise AI system with contractual data protection guarantees, or a locally deployed model with no external data transmission. The obligation to decline would not have been absolute; it would have been an obligation to find a compliant solution or to proceed without AI assistance. This counterfactual illuminates a structural point: the ethical failure was not the decision to use AI per se, but the decision to use a specific category of AI tool - open-source, publicly accessible - without first obtaining client consent for the data exposure that use necessarily entailed. Had Engineer A followed the disclosure-and-consent pathway, the subsequent work product might have been produced differently - perhaps with a privacy-compliant AI tool, perhaps without AI assistance at all - but the client relationship and the engineer's ethical standing would have been preserved. The counterfactual also suggests that the Board's conclusion that AI use is not unethical per se should be understood as conditional on the use of appropriate tools under appropriate consent frameworks, not as a blanket endorsement of any AI tool for any purpose.

The tension between Client Data Confidentiality in AI Tool Use and Mentorship Continuity and Succession Planning exposes a systemic vulnerability that the Board's conclusions do not address: Engineer A's loss of Engineer B's peer review created professional pressure to substitute AI assistance for human oversight, but the only available AI tool was open-source, meaning that satisfying the need for quality assurance necessarily required exposing Client W's confidential site data and groundwater monitoring information to a public platform without prior consent. This creates a structural conflict in which the engineer cannot simultaneously honor the confidentiality obligation and use the available compensating mechanism. The case teaches that this conflict is not resolvable by choosing one principle over the other after the fact - it is resolvable only by proactive planning before the engagement begins. The principle of Mentorship Continuity and Succession Planning, read alongside the confidentiality obligation under Code provision II.1.c, implies that when a primary quality assurance mechanism is lost, the engineer's first obligation is to identify a compliant replacement - whether a qualified peer reviewer, a privacy-compliant AI platform, or a scope limitation - before accepting work that cannot be competently and confidentially performed alone. Engineer A's failure to engage in that prior planning rendered the confidentiality breach not merely a procedural lapse but a foreseeable consequence of an inadequately structured professional practice.