Step 4: Review

Hold paramount the safety, health, and welfare of the public.

Perform services only in areas of their competence.

Avoid deceptive acts.

Engineers shall not reveal facts, data, or information without the prior consent of the client or employer except as authorized or required by law or this Code.

Engineers shall undertake assignments only when qualified by education or experience in the specific technical fields involved.

Engineers shall not affix their signatures to any plans or documents dealing with subject matter in which they lack competence, nor to any plan or document not prepared under their direction and control.

Engineers shall avoid all conduct or practice that deceives the public.

Engineers shall conform with state registration laws in the practice of engineering.

Engineers shall give credit for engineering work to those to whom credit is due, and will recognize the proprietary interests of others.

The Board cited this case to establish historical precedent for the ethical use of computer-assisted drafting and design tools, and to show that the BER has long grappled with questions about new technology in engineering practice, including early references to AI.

The Board cited this case to draw a parallel to Engineer A's use of AI-assisted drafting tools without sufficient competency verification, and to reinforce the principle that technology must not replace or substitute for engineering judgment, while also distinguishing Engineer A as not wholly incompetent unlike the engineer in that case.

Engineer A's use of AI in report writing was partly ethical, and partly unethical.

The use of AI-assisted drafting tools by Engineer A was not unethical per se.

Similar to other software used in the design or detailing process, Engineer A has no professional or ethical obligation to disclose AI use to Client W (unless such disclosure is required under Engineer A’s contract with Client W).

Beyond the Board's finding that Engineer A's use of AI in report writing was partly ethical and partly unethical, the most independently significant ethical breach was not the use of AI itself but rather Engineer A's upload of Client W's confidential site data and groundwater monitoring information into an open-source AI platform without obtaining Client W's prior consent. This act violated Code Section II.1.c regardless of the quality of the resulting report or the thoroughness of Engineer A's subsequent review. The confidentiality breach is analytically separable from the competence and disclosure questions: even if Engineer A had produced a flawless, fully attributed report, the unauthorized exposure of proprietary client data to an open-source system with unknown data retention and third-party access policies would remain an independent ethical violation. The Board's conclusion that the report submission was partly ethical should therefore be understood as resting on a foundation that was itself compromised before Engineer A wrote a single word of review commentary.

The Board's finding that AI-assisted drafting tools are not unethical per se must be qualified by a competence threshold that Engineer A failed to meet before deploying those tools on a live client engagement. Code Sections I.2 and II.2.a require engineers to undertake assignments only when qualified by education or experience in the specific technical field involved. This obligation extends to the tools an engineer selects to perform that work: adopting AI drafting software that was entirely new to the market, with which Engineer A had no prior experience, for the purpose of generating sealed engineering design documents, is not ethically equivalent to using a mature, well-understood software application. The analogy the Board drew between AI tools and conventional engineering software—such as CAD or finite element analysis programs—understates the difference in verification burden. Established engineering software produces deterministic outputs that engineers can validate against known benchmarks; large-language-model and AI-drafting outputs are probabilistic, context-sensitive, and capable of producing plausible-looking but technically incorrect results, as the design deficiencies in this case demonstrated. Engineer A's lack of tool-specific proficiency therefore created a foreseeable competence gap that the cursory design review did not close, and the Board's permissive conclusion about AI tool use should not be read to excuse adoption of unfamiliar AI tools without a minimum threshold of tool-specific qualification.

The Board's conclusion that Engineer A had no professional or ethical obligation to disclose AI use in the report—by analogy to conventional engineering software—does not adequately account for the distinct authorship and attribution dimensions raised by large-language-model text generation. Code Section III.9 requires engineers to give credit for engineering work to those to whom credit is due. When an AI system generates the substantive prose of a professional report, the question of authorship credit is not merely a stylistic preference but a professional representation about the origin of the intellectual work product. Client W's observation that the report read as if written by two different authors is not a trivial aesthetic complaint: it is evidence that the AI-generated sections were perceptibly distinct in character from Engineer A's own writing, which means the report implicitly represented a unified human authorship that did not exist. This implicit misrepresentation, even absent an affirmative false statement, is in tension with Code Sections I.5 and III.3, which prohibit deceptive acts and conduct that deceives the public. The Board's non-disclosure conclusion should therefore be treated as contingent on circumstances where AI involvement is not perceptible and does not create a false impression of authorship—conditions that were not met in this case.

The Board's partial ethical approval of Engineer A's report work—grounded in Engineer A's thorough cross-checking—should not be extended to the engineering design documents, where the standard of review was categorically different and the consequences of error were categorically more serious. For the report, Engineer A verified factual claims against journal articles and checked phrasing for originality; this is a form of substantive intellectual engagement with the AI output. For the design documents, Engineer A conducted only a cursory review and adjusted certain elements for site-specific conditions, yet affixed a professional seal. Code Section II.2.b prohibits engineers from signing plans dealing with subject matter in which they lack competence, and the act of sealing is a public representation that the engineer has exercised responsible charge over the work. A cursory review of AI-generated plans that contained misaligned dimensions and omitted safety features required by local regulations does not constitute responsible charge; it constitutes a delegation of engineering judgment to an unqualified automated system. The seal therefore misrepresented the degree of professional oversight actually applied, creating a risk to public safety that the Board correctly identified as an ethical violation but that deserves emphasis as the most serious failure in this case.

A dimension the Board did not address is the structural role that Engineer B's retirement played in creating the conditions for Engineer A's ethical failures. Engineer A's loss of mentorship and quality assurance review was not merely a personal inconvenience but a foreseeable competence gap that Engineer A was obligated to address through affirmative professional measures before accepting or continuing sealed deliverable work. The NSPE Code's competence obligations are not satisfied by the historical fact of prior mentored experience; they require that an engineer be currently qualified to perform the work independently or with appropriate oversight at the time of performance. Engineer A's response to the mentorship gap—unilaterally adopting an unfamiliar open-source AI tool as a functional substitute for peer review—was not a professionally adequate adaptation. Ethically sound alternatives would have included engaging a qualified substitute reviewer, limiting the scope of accepted work to tasks within Engineer A's independent competence, or declining the engagement pending identification of appropriate oversight. The Board's analysis would benefit from explicitly recognizing that the mentor loss was not a mitigating circumstance but a triggering condition that elevated Engineer A's professional obligations rather than relaxing them.

The Board's discretionary approach to AI disclosure—treating it as ethically favored but not categorically required absent contractual obligation—produces an unstable standard that is likely to erode client trust and professional accountability as AI tools become more capable and more widely used. From a consequentialist perspective, a universal norm requiring disclosure whenever AI plays a substantial role in generating a sealed work product would produce better outcomes across three dimensions: it would enable clients to make informed decisions about the work product they are receiving; it would create professional incentives for engineers to maintain genuine responsible charge rather than delegating judgment to AI systems; and it would allow the profession to develop empirical knowledge about where AI-assisted engineering succeeds and fails. The Board's analogy to conventional engineering software is likely to become increasingly untenable as AI systems move from computation tools to generative authorship tools, and the profession would benefit from proactively establishing disclosure norms before the absence of such norms produces more serious public safety failures than the design deficiencies observed in this case.

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 without obtaining Client W's prior consent constitutes an independent and serious breach of the client confidentiality obligation under Code Section II.1.c. Open-source AI platforms, by their nature, may retain, process, or expose input data in ways that fall outside the engineer's control, and Engineer A's unfamiliarity with the tool's full functionality compounds this risk. This breach is analytically separable from the quality of Engineer A's subsequent review of the report: even if the report were technically flawless, the act of exposing confidential client data to an uncontrolled third-party platform without consent is independently unethical. The breach does not become retroactively permissible because no confirmed data leak occurred; the obligation under II.1.c is triggered by unauthorized disclosure, not by demonstrated harm. Accordingly, the submission of the report is rendered ethically deficient on confidentiality grounds alone, independent of any assessment of report quality or AI disclosure obligations.

In response to Q102: Engineer A did not satisfy the competence requirement under Code Sections I.2 and II.2.a before adopting the AI-assisted drafting tool for a live client engagement. The tool was new to the market, Engineer A had no prior experience with it, and Engineer A's own unfamiliarity with its full functionality is explicitly acknowledged in the case facts. Competence under the Code is not limited to domain expertise in environmental engineering; it extends to the tools and methods an engineer employs to produce sealed deliverables. Using an untested, unfamiliar tool on a live engagement — particularly one generating design documents that would bear a professional seal — without any prior validation, pilot testing, or independent verification protocol falls below the standard of care the Code demands. At minimum, before relying on AI-assisted drafting software for sealed deliverables, an engineer should: (1) conduct independent validation of the tool's outputs against known correct solutions; (2) understand the tool's known failure modes and limitations; and (3) establish a verification protocol sufficient to catch errors the tool is known to produce. Engineer A did none of these things for the design documents, and the resulting deficiencies — misaligned dimensions and omitted safety features — are a foreseeable consequence of that failure.

In response to Q103: Engineer A's omission of any disclosure about AI authorship, in the context of a report that Client W independently observed 'read as if written by two different authors,' constitutes a deceptive act or misrepresentation of authorship under Code Sections I.5 and III.3, independent of whether the factual content of the report was accurate. Deception under the Code does not require an affirmative false statement; it encompasses conduct that creates a false impression in the mind of the recipient. By submitting the report under their professional seal without any attribution of AI involvement, Engineer A implicitly represented the work as their own original professional writing. The stylistic inconsistency noticed by Client W is direct evidence that the AI's contribution was perceptible and material — not merely a background tool analogous to spell-check or grammar software. When a client can detect that a work product appears to have multiple authors, and the engineer withholds the explanation that one of those apparent 'authors' is an AI system, the omission crosses from permissible non-disclosure into functional misrepresentation. This conclusion is independent of the Board's finding that AI-assisted drafting is not unethical per se; the ethical problem here is not the use of AI but the concealment of its perceptible and substantial role.

In response to Q104: Engineer A's loss of Engineer B as mentor and quality assurance reviewer created a foreseeable and material competence gap that Engineer A was ethically obligated to address through means other than the unilateral adoption of an unfamiliar AI tool. The Code's competence obligations under Sections I.2 and II.2.a do not permit an engineer to substitute an untested technology for the human oversight and peer review that previously ensured the adequacy of their work product. Engineer A had several ethically permissible alternatives: engaging a substitute peer reviewer with relevant expertise, limiting the scope of accepted work to deliverables within Engineer A's independent capacity, or declining the engagement pending identification of adequate quality assurance support. Instead, Engineer A accepted both a complex technical report and engineering design documents — deliverables that would bear a professional seal — while simultaneously losing their primary quality assurance resource and adopting an unfamiliar AI tool as a replacement. This compounded the risk rather than mitigating it. The design deficiencies that resulted — misaligned dimensions and omitted safety features — are a direct and foreseeable consequence of this failure to address the competence gap through appropriate professional means.

In response to Q201: A genuine and irresolvable tension exists between the principle of technical accuracy achieved through thorough review and the principle of honest representation of authorship and process. The Board's finding that Engineer A's thorough cross-checking of the report satisfied the competence obligation addresses only the output quality dimension of the ethical analysis. It does not resolve — and in fact obscures — the separate question of whether Engineer A misrepresented the nature and origin of the work product. An engineer may produce a technically accurate report while simultaneously creating a false impression about who or what authored it. When these two obligations cannot be simultaneously fulfilled — that is, when using AI produces a technically adequate product but concealing AI use constitutes deception — the obligation of candor and non-deception should take precedence, because the Code's prohibitions on deceptive acts under Sections I.5 and III.3 are categorical in character and do not yield to consequentialist justifications based on output quality. Technical accuracy is a necessary but not sufficient condition for ethical compliance.

In response to Q203: A fundamental conflict exists between the act of affixing a professional seal and the adequacy of the oversight that seal is intended to certify. The professional seal is not merely a formality; under Code Section II.2.b and state engineering seal law, it represents the engineer's certification that they exercised responsible charge over the work. Engineer A's cursory review of AI-generated design documents — which failed to detect misaligned dimensions and omitted safety features required by local regulations — means that the seal was affixed to documents that Engineer A had not adequately reviewed. The seal therefore functioned not as a certification of responsible charge but as a misrepresentation of it. This is among the most serious ethical failures in the case, because the seal is the primary mechanism by which the public and clients rely on the engineer's professional judgment. To prevent this failure mode, procedural safeguards for sealing AI-generated design documents should include: (1) a systematic, element-by-element verification of all dimensions and specifications against source data; (2) explicit cross-checking of all design features against applicable local, state, and federal regulatory requirements; and (3) documented evidence of the review process sufficient to demonstrate responsible charge. A cursory review is categorically insufficient when AI has generated the primary content of sealed design documents.

In response to Q301 and Q302: From a deontological perspective, Engineer A failed two categorical duties simultaneously. First, the duty of candor to Client W was breached by omitting disclosure of AI involvement in both the report and the design documents. A Kantian analysis requires asking whether the maxim 'engineers may omit disclosure of AI authorship when the output is technically adequate' could be universalized without contradiction. It cannot: if all engineers concealed AI involvement whenever they judged the output satisfactory, clients would be systematically deprived of information material to their ability to evaluate the work product and the professional relationship, undermining the very trust on which professional licensure depends. Second, the duty to protect client confidentiality was breached by uploading Client W's proprietary data to an open-source platform without consent. This duty is categorical under Code Section II.1.c and does not admit of exceptions based on the engineer's assessment of likely harm. The absence of a confirmed data breach is irrelevant to the deontological analysis: the wrong consists in the unauthorized act of disclosure itself, not in its consequences. Both violations are independently sufficient to render Engineer A's conduct unethical under a deontological framework, and they compound each other in a way that reflects a broader pattern of prioritizing personal convenience over professional obligation.

In response to Q303: From a consequentialist perspective, the aggregate outcomes of Engineer A's AI-assisted workflow produce a net harm that substantially outweighs the efficiency gains sought. The benefits were limited to time savings in drafting and the ability to meet deliverable deadlines without a quality assurance reviewer. The harms include: (1) exposure of Client W's confidential site and groundwater data to an uncontrolled open-source platform, creating ongoing and unquantifiable data security risk; (2) submission of design documents containing misaligned dimensions and omitted safety features required by local regulations, creating direct public safety risk and requiring costly revision; (3) erosion of Client W's trust, evidenced by the client's observation that the report appeared to have two authors and by the client's instruction to revise the design documents; and (4) the broader systemic harm of normalizing inadequate AI oversight in professional engineering practice. The efficiency gain from AI adoption is only ethically defensible when the engineer has sufficient familiarity with the tool to ensure that the output meets professional standards. Here, Engineer A's unfamiliarity with the tool, combined with a cursory review of the design documents, meant that the efficiency gain was achieved by transferring risk to the client and the public rather than by genuinely improving productivity.

In response to Q304 and Q305: From a virtue ethics perspective, Engineer A's conduct in both the report and design document contexts falls short of the professional integrity and responsible stewardship that the public and clients are entitled to expect from a licensed engineer. A virtuous engineer, confronted with the loss of a trusted mentor and quality assurance reviewer, would respond by seeking alternative qualified oversight — not by substituting an unfamiliar, unvalidated AI tool for human professional judgment. The choice to adopt an open-source AI drafting tool with no prior experience, on a live client engagement involving sealed deliverables, reflects a disposition toward self-reliance that is admirable in principle but reckless in execution when the engineer lacks the competence to validate the tool's outputs. More seriously, Engineer A's decision to affix a professional seal to AI-generated design documents after only a cursory review reflects a self-serving shortcut that undermines the moral foundation of professional licensure. The seal exists to assure the public that a qualified professional has exercised genuine judgment over the work. Affixing it to documents that Engineer A had not adequately reviewed — and that contained detectable errors — is not merely a procedural lapse; it is a betrayal of the public trust that the seal is designed to embody. A virtuous engineer would have either declined to seal the documents until a thorough review was complete or sought qualified assistance before submission.

In response to Q306: A universal norm requiring engineers to disclose AI involvement whenever AI plays a substantial role in generating a work product would produce better outcomes for public safety, client trust, and the engineering profession than the current discretionary approach endorsed by the Board. The Board's analogy between AI tools and conventional engineering software — such as CAD or structural analysis programs — is imperfect in a critical respect: conventional engineering software executes deterministic calculations based on engineer-defined inputs, while large language model AI generates probabilistic text and design content that may contain errors, hallucinations, or stylistic artifacts that are not traceable to any specific engineer input. This fundamental difference in the nature of the tool's contribution justifies a different disclosure standard. Mandatory disclosure of substantial AI involvement would: (1) enable clients to make informed decisions about the work product and the professional relationship; (2) create accountability incentives that encourage engineers to conduct more rigorous reviews of AI-generated content; (3) allow the profession to develop empirical data about AI tool reliability across different application domains; and (4) prevent the kind of functional deception illustrated by this case, where a client independently detected AI involvement through stylistic inconsistency. The current discretionary approach, by contrast, permits concealment of AI involvement as long as the engineer judges the output to be adequate — a standard that is self-assessed, unverifiable, and inconsistent with the Code's broader commitment to transparency and public trust.

In response to Q401: Prior disclosure to Client W and explicit written consent for uploading confidential site data to an open-source AI platform would have resolved the confidentiality breach under Code Section II.1.c and would have substantially altered the ethical assessment of the engagement. Had Engineer A disclosed the intended use of open-source AI software before beginning work — including the fact that confidential client data would be input into the platform — Client W would have had the opportunity to evaluate the data security implications, negotiate alternative arrangements, or withhold consent. This would have either eliminated the confidentiality violation (if consent were granted) or prevented the data exposure entirely (if consent were withheld). However, prior consent for data upload would not have resolved all ethical violations: the design document deficiencies arising from inadequate review and the responsible charge failures would have persisted regardless of disclosure. The counterfactual therefore supports the conclusion that disclosure and consent are necessary but not sufficient conditions for ethical compliance — they address the confidentiality and transparency dimensions of the analysis but do not substitute for the competence and responsible charge obligations that Engineer A independently failed to satisfy.

In response to Q402: Had Engineer A engaged a qualified substitute reviewer to replace Engineer B's mentorship and quality assurance role before submitting either the report or the design documents, the design deficiencies, regulatory omissions, and responsible charge failures identified in this case would very likely have been prevented. A competent peer reviewer would have been expected to detect misaligned dimensions and omitted safety features — errors that Client W, a non-engineer client, was able to identify upon review. The presence of qualified peer review would also have satisfied the spirit of the responsible charge obligation under Code Section II.2.b, because the engineer's judgment would have been supplemented — rather than replaced — by another qualified professional. Under those circumstances, AI use would have remained ethically permissible: the tool would have served its appropriate function as a drafting aid subject to meaningful professional oversight, rather than as a substitute for the professional judgment that Engineer A lacked the independent capacity to fully exercise. This counterfactual reinforces the conclusion that the ethical failures in this case are not primarily attributable to AI use per se, but to the absence of adequate professional oversight — a gap that Engineer A was obligated to fill through qualified human review rather than through technological substitution.

In response to Q403: If Engineer A had applied the same level of thorough, cross-referenced review to the AI-generated design documents as they applied to the AI-generated report — verifying all dimensions against source data, checking safety feature requirements against local regulations, and reconciling site-specific conditions — the Board's finding of an ethical violation for failure to maintain responsible charge over the design documents would not have been warranted on review-adequacy grounds. The Board's distinction between the report (ethical) and the design documents (unethical) rests substantially on the difference in review quality: thorough review for the report, cursory review for the design documents. A thorough review of the design documents that successfully identified and corrected all misaligned dimensions and omitted safety features before submission would have satisfied the responsible charge obligation under Code Section II.2.b. However, this counterfactual does not eliminate all ethical concerns: the confidentiality breach from uploading client data to an open-source platform and the non-disclosure of AI involvement would have persisted regardless of review quality. The counterfactual therefore isolates the responsible charge violation as a curable deficiency — one that adequate review could have remedied — while confirming that the confidentiality and transparency violations are structural features of Engineer A's approach that no amount of post-generation review can retroactively cure.

In response to Q404: Voluntary disclosure of AI tool use in both the report and the design documents at the time of submission would have materially altered the client's trust, the professional relationship, and the Board's ethical assessment regarding the disclosure obligation. Client W's observation that the report 'read as if written by two different authors' demonstrates that the absence of disclosure was already functionally deceptive: the client perceived an anomaly that had a specific explanation — AI authorship of the introduction — that Engineer A withheld. Had Engineer A proactively disclosed the AI tools used and the extent of AI involvement, Client W would have had a factual explanation for the stylistic inconsistency, would have been positioned to make an informed judgment about the adequacy of Engineer A's review, and would have retained the ability to request additional assurances or alternative approaches. Such disclosure would have aligned Engineer A's conduct with the Code's requirements of candor under Sections I.5 and III.3 and would have transformed the AI use from a concealed practice into a transparent professional choice subject to client evaluation. The Board's conclusion that disclosure is not ethically required as a general matter does not address the specific circumstance where non-disclosure is already functionally deceptive — a circumstance this case presents directly. In that specific circumstance, the ethical obligation to avoid deception independently requires disclosure, regardless of whether a general disclosure norm exists.

The most consequential unresolved tension in this case is between Engineer A AI Report Competence and Engineer A AI Authorship Concealment. The Board found that Engineer A's thorough cross-checking of the report satisfied the competence standard, effectively decoupling technical accuracy from the question of authorship transparency. However, this resolution is incomplete: an engineer can produce a factually correct work product while simultaneously misrepresenting—by omission—the nature and origin of that product. Client W's observation that the report read as if written by two different authors demonstrates that the concealment was not merely theoretical; it was perceptible and materially affected the client's understanding of what they received. The Board's implicit prioritization of output quality over process transparency teaches that, under the current framework, competence obligations can be satisfied independently of candor obligations. This case suggests that framework is insufficient: where AI plays a substantial authorship role and that role is perceptible to the client, the principle of Engineer A AI Disclosure Transparency should be treated as a corollary of competence rather than a separate, discretionary consideration. Competence and candor should be jointly necessary conditions for ethical submission of a sealed work product.

The tension between Engineer A Responsible Charge Design Failure and Engineer A Seal on AI Design Documents exposes a structural vulnerability in how professional sealing norms interact with AI-generated outputs. The professional seal is the legal and ethical mechanism by which an engineer certifies responsible charge—meaning the engineer has exercised sufficient judgment, direction, and verification over the work product to stand behind it publicly. In this case, Engineer A's cursory review of AI-generated design documents that contained misaligned dimensions and omitted safety features required by local regulations means the seal was affixed to a product over which responsible charge had not, in fact, been exercised. The seal did not confirm oversight; it obscured its absence. This tension was not resolved by the Board so much as it was exposed: the Board's finding of an ethical violation for the design documents implicitly acknowledges that sealing AI-generated work after only a high-level review fails the responsible charge standard. The case teaches that the principle of Engineer A Seal on AI Design Documents cannot be treated as equivalent to sealing conventionally produced documents unless the engineer's review is substantively equivalent in depth and rigor to what would have been required had the engineer produced the documents without AI assistance. The seal's moral and legal weight demands that AI-assisted design receive verification proportionate to the risk of the output, not merely a review proportionate to the engineer's familiarity with the tool.

The interaction among Engineer A Client Data AI Upload, Engineer A Professional Competence AI Use, and Engineer A Mentor Loss Response reveals that what appears on the surface to be a single decision—adopting an AI tool—was in fact a compounded ethical failure involving three independently assessable breaches that reinforced one another. First, uploading Client W's confidential site data and groundwater monitoring information into an open-source AI platform without prior consent violated the client confidentiality principle regardless of whether any data breach materialized, because the obligation under Code Section II.1.c is triggered by unauthorized disclosure, not by demonstrated harm. Second, adopting an AI drafting tool that was entirely new to the market with no prior experience, and applying it to sealed deliverables for a live client engagement, violated the competence principle because Engineer A could not evaluate the tool's accuracy, limitations, or failure modes before relying on it. Third, Engineer A's response to the loss of Engineer B as mentor and quality assurance reviewer—substituting an unfamiliar AI tool rather than seeking a qualified peer reviewer, limiting scope, or declining work beyond independent capacity—violated the principle of Engineer A Mentor Loss Response, which required Engineer A to address the competence gap through means that preserved rather than compounded professional risk. The case teaches that when multiple principles are simultaneously in tension, the ethical analysis must treat each breach as independently actionable rather than allowing a finding of compliance on one dimension—such as the quality of the report review—to offset or absorb violations on others. Engineer A AI Tool Non-Disclosure Report and Engineer A AI Disclosure Transparency cannot be resolved by analogy to conventional software tools when the tool in question is unfamiliar, open-source, and processing confidential client data, because that analogy assumes a baseline of tool competence and data security that was absent here.

Was Engineer A’s use of AI to create the report text ethical, given that Engineer A thoroughly checked the report?

Was Engineer A’s use of AI-assisted drafting tools to create the engineering design documents ethical, given that Engineer A reviewed the design at a high level?

If the use of AI was acceptable, did Engineer A have an ethical obligation to disclose the use of AI in any form to the Client?

By uploading Client W's confidential site data and groundwater monitoring information into an open-source AI platform without obtaining Client W's prior consent, did Engineer A breach the client confidentiality obligation under Code Section II.1.c, and does that breach independently render the report submission unethical regardless of the quality of Engineer A's subsequent review?

Given that Engineer A used an AI drafting tool that was entirely new to the market and with which Engineer A had no prior experience, did Engineer A satisfy the competence requirement under Code Sections I.2 and II.2.a before adopting that tool for a live client engagement, and what minimum level of tool-specific proficiency should be required before an engineer may rely on AI-assisted drafting software for sealed deliverables?

Because the AI-generated report produced stylistically inconsistent text that Client W noticed and attributed to what appeared to be two different authors, did Engineer A's omission of any disclosure about AI authorship constitute a deceptive act or misrepresentation of authorship under Code Sections I.5 and III.3, independent of whether the factual content of the report was accurate?

To what extent did Engineer A's loss of Engineer B as a mentor and quality assurance reviewer create a foreseeable competence gap that Engineer A was obligated to address through means other than unilateral adoption of an unfamiliar AI tool—such as engaging a substitute peer reviewer, limiting the scope of accepted work, or declining the engagement—before undertaking sealed deliverables for Client W?

Does the principle of Engineer A AI Report Competence—which the Board found satisfied by Engineer A's thorough cross-checking of the report—conflict with the principle of Engineer A AI Authorship Concealment, in that an engineer may produce a technically accurate work product while simultaneously misrepresenting the nature and origin of that work product, and if so, which obligation should take precedence when the two cannot be simultaneously fulfilled?

Does the principle of Engineer A AI Disclosure Transparency—which the Board acknowledged is ethically favored when AI plays a substantial role—conflict with the principle of Engineer A AI Tool Non-Disclosure Report, which the Board grounded in an analogy to conventional engineering software that carries no mandatory disclosure obligation, and how should engineers resolve this tension in the absence of a universal regulatory standard?

Does the principle of Engineer A Responsible Charge Design Failure—arising from Engineer A's cursory review of AI-generated plans that contained misaligned dimensions and omitted safety features—conflict with the principle of Engineer A Seal on AI Design Documents, in that affixing a professional seal is intended to certify responsible charge but the act of sealing here obscured rather than confirmed the adequacy of engineering oversight, and what procedural safeguards should govern the sealing of AI-generated design documents?

Does the principle of Engineer A Client Data AI Upload—which reflects the engineer's interest in leveraging available tools to meet client deliverable demands—conflict with the principle of Engineer A Professional Competence AI Use, in that using an unfamiliar open-source AI tool to process confidential client data may simultaneously undermine both the client's data security interests and the engineer's own competence obligations, and should the ethical analysis treat these as a single compounded violation or as two independently assessable breaches?

From a deontological perspective, did Engineer A fulfill their duty of candor to Client W by omitting any disclosure of AI involvement in both the report and the engineering design documents, regardless of whether that omission ultimately harmed the client?

From a deontological perspective, did Engineer A violate a categorical duty to protect client confidentiality by uploading Client W's proprietary site data and groundwater monitoring information into an open-source AI platform without obtaining prior consent, irrespective of whether any actual data breach occurred?

From a consequentialist perspective, did the aggregate outcomes of Engineer A's AI-assisted workflow — including the polished but stylistically inconsistent report, the deficient design documents with misaligned dimensions and omitted safety features, and the potential exposure of confidential client data — produce a net harm that outweighs the efficiency gains Engineer A sought by adopting AI tools?

From a virtue ethics perspective, did Engineer A demonstrate the professional integrity and intellectual honesty expected of a licensed engineer by choosing to use an unfamiliar, open-source AI drafting tool — with no prior experience — as a substitute for the mentorship and quality assurance oversight previously provided by Engineer B, rather than seeking alternative qualified review or declining work beyond their current independent capacity?

From a virtue ethics perspective, does Engineer A's decision to apply their professional seal to AI-generated design documents after only a cursory review reflect the character trait of responsible stewardship that the public and clients are entitled to expect from a licensed professional engineer, or does it represent a self-serving shortcut that undermines the moral foundation of professional licensure?

From a consequentialist perspective, would a universal norm requiring engineers to disclose AI involvement whenever AI plays a substantial role in generating a work product produce better outcomes for public safety, client trust, and the engineering profession than the current discretionary approach endorsed by the Board?

If Engineer A had disclosed to Client W, prior to beginning work, that they intended to use open-source AI software and had obtained explicit written consent for uploading confidential site data to that platform, would the ethical violations identified by the Board regarding client data exposure and lack of attribution have been avoided, and would the overall ethical assessment of the engagement have changed?

What if Engineer A had sought a qualified substitute reviewer — such as a peer engineer or a licensed colleague — to replace Engineer B's mentorship and quality assurance role before submitting either the report or the design documents? Would the design deficiencies, regulatory omissions, and responsible charge failures identified by the Board have been prevented, and would AI use have remained ethically permissible under those circumstances?

If Engineer A had conducted the same level of thorough, cross-referenced review on the AI-generated design documents as they applied to the AI-generated report — verifying dimensions, checking safety feature requirements against local regulations, and reconciling site-specific conditions — would the Board's finding of an ethical violation for failure to maintain Responsible Charge over the design documents still have been warranted?

What if Engineer A had voluntarily disclosed the use of AI tools in both the report and the design documents — including the specific AI software used and the extent of AI involvement — at the time of submission to Client W? Would such proactive transparency have altered the client's trust, the professional relationship, or the Board's ethical assessment regarding the disclosure obligation, and does the client's observation that the report 'read as if written by two different authors' suggest that non-disclosure was already functionally deceptive?

Accepting the project within the bounds of competence matters because it establishes the foundational professional commitment to public safety, and any downstream failures in design review or AI tool use trace back to whether the engineer had the genuine capability to oversee the full scope of work from the outset.

Adopting the AI tool for the report fulfils competence obligations by leveraging available technology, but the causal chain shows it directly produces an AI Disclosure Omission, meaning the fulfillment of competence is undermined by a transparency failure that leaves the client and public unaware of how the deliverable was generated.

Using AI for design generation under dual deliverable pressure carries normative weight because the causal chain produces AI-generated design outputs that feed into a review process already shown to be cursory, meaning the absence of a clear fulfillment or violation here reflects a gap in oversight that amplifies the risk of undetected design defects reaching the client.

Uploading client data to the AI tool violates confidentiality because it causally produces client data exposure to a third-party system, and this breach is not offset by the downstream report generation benefit, since the client never consented to their proprietary information leaving the engineer's control.

Conducting a thorough review fulfils both direction and control over the work product and competence obligations, and this matters causally because the report stylistic inconsistency produced by AI generation and draft sealing could have been caught and corrected before submission, making the review the critical intervention point for maintaining professional accountability.

Because the AI tool adoption directly caused the report generation and its stylistic inconsistencies, concealing the AI's role deceives the client about the true authorship and quality basis of the deliverable, which undermines the trust that professional attribution obligations are designed to protect.

Sealing and submitting the report gives it the force of a professional certification, so violating credit and attribution obligations at this step matters because it formally commits Engineer A to standing behind a product whose actual origins were misrepresented, compounding the downstream stylistic inconsistency into a documented professional record.

The cursory review caused a design defect to be discovered only after submission, which triggered a client revision instruction, meaning the failure to exercise adequate direction and control over the work product directly produced a safety-relevant gap that the responsible charge obligation exists precisely to prevent.

Submitting the design document without adequate review or qualification fulfils no protective obligation and instead places a potentially defective design into the project record, meaning the public safety and responsible charge violations are not merely procedural but carry real downstream consequences given that a defect was in fact present.

This question arose because the act of uploading client data to an open-source AI platform is a factual trigger that sits at the intersection of two legitimate professional obligations, one demanding confidentiality and one demanding competent direction of tools. The deontological framing sharpens the question by asking whether the confidentiality duty is categorical and outcome-independent, which forces a determination of whether the absence of an actual breach is legally or ethically irrelevant to the duty itself.

This question arose because Engineer A's AI-assisted workflow produced multiple distinct categories of harm across a single project, and consequentialism demands a unified net assessment rather than isolated evaluation of each failure. The question became necessary because the data shows real efficiency motivation alongside real harm production, and no single warrant resolves whether the aggregate outcome is net negative without contested empirical judgments about harm probability, severity, and the counterfactual value of the efficiency gains.

This question emerged because Engineer A sealed and submitted AI-generated work products to Client W without any acknowledgment of AI involvement in either deliverable, creating a factual record of non-disclosure that directly contests the deontological principle that candor to a client is an obligation owed regardless of consequences. The question is not resolved by pointing to outcome because deontological analysis asks whether the duty was honored at the moment of action, and the competing warrants disagree on whether non-disclosure of AI authorship constitutes a breach of that duty or merely a lapse in attribution practice.

This question arose because the data simultaneously supports two defensible positions: that AI tools are legitimate engineering aids when an engineer seals and takes responsibility for the work, and that a high-level review of AI-generated design documents falls short of the direction and control required by responsible charge. The Engineer A Design Deficiencies state, combined with the Engineer A Unfamiliar Drafting Tool state and the loss of mentorship oversight captured in Engineer A Mentor Unavailable, created a factual record that makes it genuinely contestable whether Engineer A's review process met the standard that would make the AI tool use ethical.

The question emerged because Engineer A produced client deliverables substantially through AI tools and then omitted any mention of that fact, placing the client in a position of receiving work without knowing its origin. The absence of an explicit disclosure rule created genuine uncertainty about whether the obligation to be transparent about authorship and method extends beyond the engineer's internal professional responsibility and reaches the client relationship directly.

The question arose because Engineer A's thorough review satisfies the professional competence and responsible charge obligations that normally govern work product quality, but the act of submitting AI-generated text without disclosure leaves the attribution and intellectual honesty obligations unresolved. The two sets of obligations point to different conclusions about the same conduct, and neither fully overrides the other given the facts, which is precisely what generates the ethical question.

This question arose because Engineer A's decision to input Client W's confidential data into an open-source AI platform created a potential confidentiality violation that is analytically separate from any question about the report's technical adequacy. The question forces a determination of whether a procedural breach in how work was generated can independently condemn a submission that might otherwise be technically sound, which the Code's structure does not explicitly resolve.

The question emerged because Engineer A's state of unfamiliarity with the AI drafting tool, combined with the loss of mentor oversight after Supervisor Retirement and the pressure of Dual Deliverable Pressure, produced sealed design documents with safety-relevant deficiencies. The absence of any profession-wide threshold for AI tool proficiency before sealed use left the competence standard contested, forcing the question of what minimum readiness is required before an engineer may rely on such tools for deliverables that carry a professional seal.

This question arose because Client W's direct observation of stylistic inconsistency transformed what might otherwise be a silent omission into a situation where the client was actively forming an incorrect inference about authorship, making Engineer A's silence function more like a tacit false assertion. The question is whether the combination of a detectable anomaly and continued non-disclosure crosses the line from mere omission into misrepresentation under the cited code sections, independent of factual accuracy.

This question arose because Engineer B's retirement created a foreseeable structural change in Engineer A's quality assurance capacity at the same moment Engineer A adopted an unfamiliar AI tool and accepted sealed deliverables, producing a compound competence deficit that the responsible charge obligation alone may not have been designed to address. The question forces analysis of whether the obligation to practice within competence required Engineer A to act before the engagement rather than only during it, and what specific remediation steps that obligation demanded.

This question emerged because the Board's competence finding and the concealment finding were reached independently, leaving open whether satisfying one obligation can offset a violation of the other. The two principles address different dimensions of professional conduct, technical quality and representational honesty, and the case provides no hierarchy for resolving their conflict when a single submission implicates both.

The question emerged because Engineer A's conduct simultaneously satisfied the factual conditions that activate both warrants: the AI contribution was substantial enough that stylistic inconsistency became detectable, yet the Board's own reasoning acknowledged the software analogy as a legitimate basis for non-disclosure. The absence of a governing regulatory standard left both warrants in play without a rule to subordinate one to the other, making the tension a genuine unresolved conflict rather than a clear violation.

This question emerged because the professional seal is designed to function as a guarantee of responsible charge, but Engineer A's cursory review of AI-generated plans containing misaligned dimensions and omitted safety features meant the seal communicated a level of oversight that had not actually occurred. The gap between what the seal formally represents and what Engineer A actually performed created a direct conflict between the obligation to seal work under responsible charge and the obligation to ensure that responsible charge was genuinely exercised before sealing.

This question emerged because Engineer A's single act of uploading confidential client data to an unfamiliar AI tool simultaneously activated two distinct ethical warrants, one protecting client privacy and one requiring professional competence, and the factual record does not resolve whether those warrants operate on the same normative ground or on independent grounds. The question of whether to treat the violations as compounded or independent is not merely taxonomic but determines how responsibility is assigned, what remedies are appropriate, and whether the ethical analysis focuses on harm to the client, harm to professional standards, or both.

This question emerged because Engineer B's retirement created a genuine competence gap at the same moment Engineer A faced dual deliverable pressure, and Engineer A responded by deploying an unfamiliar AI drafting tool rather than seeking alternative qualified oversight or declining work beyond independent capacity. The virtue ethics framing sharpens the question by asking not just whether the outcome was harmful but whether Engineer A demonstrated the self-knowledge and professional honesty that integrity requires before accepting and executing work under those conditions.

This question arose because the professional seal is the institutional symbol of responsible stewardship, and Engineer A's cursory review created a direct contest between the virtue of diligence that the seal represents and the self-serving convenience of accepting AI output without the verification that would make the seal honest. The discovery of design deficiencies after submission made the gap between the character trait the seal implies and the character trait Engineer A actually demonstrated impossible to ignore.

This question emerged because the actual harms produced by Engineer A's undisclosed AI use, including design defects, public safety risks, and client data exposure, created observable evidence that discretionary disclosure failed to protect the interests it was meant to protect. The gap between the outcomes the current approach produced and the outcomes a universal norm might have produced forced a consequentialist evaluation of whether the warrant authorizing discretion is strong enough to survive the rebuttal that mandatory disclosure would have prevented those harms.

This question arose because the Board identified multiple distinct violations, and it is genuinely unclear whether a single procedural act, prior written consent, would have collapsed those violations into a compliant engagement or whether the responsible charge, verification, and public safety failures would have remained as independent grounds for ethical censure. The question forces a structural analysis of whether the violations are causally linked to the absence of consent or whether they reflect deeper competence and oversight failures that consent alone could not have cured.

This question arose because the Board's findings identified failures across multiple dimensions simultaneously, including design deficiencies, regulatory omissions, and responsible charge deficits, and it became unclear whether those failures were caused by the absence of a substitute reviewer specifically or by a broader pattern of inadequate verification that a substitute reviewer might not have corrected. The question forces a structural analysis of whether the ethical permissibility of AI use depends on the presence of a qualified human reviewer in the oversight chain, or whether it depends on the competence and rigor of whoever holds responsible charge.

This question emerged because the Board's finding rested on the inadequacy of Engineer A's review of the design documents, but Engineer A had demonstrated the capacity for thorough AI output review in the context of the report. The question asks whether the ethical violation was a function of the review process Engineer A chose to apply, rather than an inherent consequence of using AI-assisted drafting tools, which creates genuine uncertainty about whether the violation was avoidable through conduct rather than unavoidable through context.

This question arose because the stylistic inconsistency noticed by Client W created observable evidence that the work product had a non-uniform origin, which transformed what might have been a private tool-use decision into a situation where the client was already drawing inferences about authorship without accurate information. The question then asks whether voluntary disclosure would have reframed the entire ethical situation, because if transparency at submission would have resolved the client's concern and altered the Board's analysis, then the omission itself becomes the operative ethical failure rather than the AI use.

Because Engineer A exposed Client W's confidential data to a system with unknown retention and access policies before obtaining any consent, the board found an independent breach of II.1.c that the quality of the resulting report could not retroactively cure, since the violation was complete at the moment of upload.

Because Engineer A had no experience with the AI tool and because the tool's probabilistic output characteristics created a verification burden that conventional software does not, the board concluded that adopting it for live sealed work without any tool-specific qualification failed the competence threshold under I.2 and II.2.a, regardless of whether AI tools are permissible in principle.

Because Client W could detect the stylistic discontinuity between AI-generated and human-written sections, Engineer A's silence about AI involvement functioned as an implicit claim of sole human authorship, placing the omission in tension with the deception prohibitions in I.5 and III.3 and the attribution obligation in III.9.

Because the design review was cursory rather than substantive, and because the sealed documents contained errors that a competent review would have caught, the board found that the seal misrepresented the degree of professional oversight applied and that this failure, given the public safety consequences of deficient design documents, was the most serious ethical violation in the case.

Because Engineer B's retirement removed a structural component of Engineer A's quality assurance process and because Engineer A's substitution of an unfamiliar AI tool for that human oversight was not a professionally adequate adaptation, the board concluded that the mentor loss elevated rather than relaxed Engineer A's obligations and that the ethical failures in this case were foreseeable consequences of failing to address that gap through appropriate professional measures.

Given that no mandatory disclosure rule currently exists and the Board analogized AI tools to conventional engineering software, the Board concluded that disclosure is ethically favored but not required, while simultaneously warning that this standard is unstable and will likely produce worse outcomes as AI capabilities grow.

Because Engineer A exposed Client W's confidential data to an uncontrolled platform without consent, the Board found an independent ethical breach under II.1.c, treating the absence of demonstrated harm as irrelevant to whether the obligation was violated.

Because Engineer A used an untested, unfamiliar tool on a live engagement without any prior validation or verification protocol, and because the resulting deficiencies were foreseeable, the Board concluded that the competence requirement under I.2 and II.2.a was not satisfied before the tool was adopted.

Because Client W could perceive the stylistic inconsistency caused by AI authorship and Engineer A withheld the explanation, the Board concluded that the omission crossed from mere non-disclosure into functional deception, independently triggering the candor obligations under I.5 and III.3 in this specific circumstance.

Because the Board found the report technically competent but simultaneously found the concealment of AI authorship perceptible and materially misleading, the Board concluded that the existing framework is insufficient and that competence and candor should be treated as jointly necessary conditions for ethical submission of a sealed work product where AI plays a substantial authorship role.

Because Engineer A reviewed the AI-generated design documents only at a high level and the documents contained misaligned dimensions and omitted required safety features, the board found that the act of sealing did not confirm responsible charge but instead obscured its absence, making the sealing and submission of those documents an ethical violation under the responsible charge standard.

Because Engineer A's single decision to adopt an AI tool compounded three distinct ethical failures that each independently violated a code obligation, the board concluded that the ethical analysis could not treat these as a unified or offsetting set of considerations, and that each breach remained actionable on its own terms regardless of Engineer A's performance on other dimensions of the engagement.

Because Engineer A thoroughly checked the AI-generated report and verified its technical content before submission, the board found that the use of AI for report drafting was not unethical in that respect, though the conclusion is partial because other aspects of the same engagement, including the data upload and the design documents, were found to be unethical.

Because the board found that AI-assisted drafting is not categorically different from other engineering software tools when the engineer exercises adequate oversight, it concluded that Engineer A's use of AI for drafting was not unethical per se, with the ethical assessment turning on the quality of the review performed rather than on the fact of AI use itself.

Because no contractual or regulatory requirement compelled disclosure and the board analogized AI drafting tools to conventional software, it concluded that Engineer A had no independent ethical obligation to disclose AI use to Client W, though this resolution is conditional on the absence of a disclosure requirement and does not address whether the client-observable stylistic inconsistency independently triggered a duty to correct a potentially deceptive impression.

Because Client W could detect a stylistic discontinuity consistent with multiple authors, and because Engineer A offered no explanation, the board found that the omission created a false impression of sole human authorship. The conclusion rests on the perceptibility of the AI contribution in this case, not on a universal rule that all AI use must be disclosed.

Because Engineer A simultaneously lost their mentor, accepted complex sealed deliverables, and adopted an unfamiliar AI tool without any substitute peer review, the board found that the competence gap was foreseeable and the resulting design deficiencies were a direct consequence of failing to address it through appropriate professional means.

Because Engineer A's thorough review addressed only output quality and left the authorship misrepresentation intact, the board found that satisfying the competence obligation does not discharge the separate and categorical obligation of candor, and that the latter must prevail when the two cannot be simultaneously fulfilled.

Because Engineer A's review did not detect errors that a systematic check would have caught, the board found that the seal functioned as a misrepresentation of responsible charge rather than a certification of it, and prescribed specific procedural safeguards to prevent this failure mode when AI generates sealed design content.

Because Engineer A acted without client consent on the data upload and without disclosure on the authorship question, the board found two independent categorical violations that compound each other as a pattern of prioritizing personal convenience over professional obligation, with the absence of a confirmed breach being irrelevant to the deontological analysis.

Because Engineer A was unfamiliar with the tool and reviewed the design documents only at a high level, the efficiency gains were achieved by offloading risk onto the client and the public rather than by genuinely improving productivity, and the board found that the aggregate harms across data security, public safety, client trust, and professional norms substantially outweighed those gains.

Because Engineer A chose technological substitution over qualified human oversight after losing Engineer B, and then sealed documents containing detectable errors following only a cursory review, the board found that both decisions reflected a disposition that fell short of the professional integrity and responsible stewardship that licensure demands.

Because AI language models produce content through probabilistic processes that differ fundamentally from deterministic engineering software, and because non-disclosure in this case was already functionally deceptive as evidenced by the client's independent detection of AI involvement, the board concluded that a universal disclosure norm for substantial AI involvement would produce better outcomes than the current discretionary standard.

Because Engineer A's violations included both the unauthorized data upload and the independent failures of competence and responsible charge over the design documents, the board found that prior consent would have eliminated the confidentiality breach but would have left the design deficiency and sealing violations intact.

Because the design deficiencies were detectable by a non-engineer client and arose from the absence of qualified oversight rather than from AI use per se, the board concluded that engaging a qualified substitute reviewer before submission would have prevented the responsible charge failures and preserved the ethical permissibility of AI-assisted drafting.

Given that the Board's distinction between the report and the design documents rested substantially on the difference in review depth rather than on the mere fact of AI use, the Board concluded that a counterfactually thorough review of the design documents would have satisfied the responsible charge obligation under II.2.b, because the deficiencies in the submitted documents were curable through the same verification process Engineer A successfully applied to the report. However, because the confidentiality breach and the non-disclosure of AI involvement were structural features of Engineer A's workflow that preceded and were independent of any review step, those violations would have persisted even under the counterfactual, confirming that responsible charge is a curable deficiency while confidentiality and transparency violations are not.

Should Engineer A (and Engineer B as mentor) proactively disclose to Client W that AI tools were used to generate portions of the report and design, or treat AI as an internal drafting tool requiring no separate client notification?

Should Engineer A refrain from uploading Client W's confidential site and groundwater data into open-source AI platforms, or is use of such platforms permissible when the engineer exercises professional judgment over the outputs?

Must Engineer A verify, independently direct, and properly attribute all AI-generated technical content in sealed reports and design documents, or is applying standard QA review to AI outputs sufficient to satisfy responsible charge obligations?

Should Engineer A proactively disclose to the client that AI tools generated substantial portions of the design and report, or present the deliverables as conventionally authored engineering work?

Should Engineer B require Engineer A to disclose AI authorship of the deliverables to the client, or accept the work as submitted on the basis that Engineer B's review satisfies the professional responsibility?

Should Engineer A seal the AI-generated design documents based on the review conducted, or withhold the seal until a more rigorous independent verification of safety compliance and responsible charge is completed?

Should Engineer B conduct a full independent technical review of the AI-assisted design documents before sealing, or apply a standard supervisory review comparable to reviewing conventional CAD-produced work?

Should Engineer A proactively disclose to the client and to Engineer B the nature and extent of AI tool usage in the design and report before those documents are finalized, or treat the AI tools as internal drafting aids that do not require separate disclosure?

Should Engineer A seal and submit the AI-generated design documents after only a cursory review, or must Engineer A conduct a thorough independent technical verification before sealing?

Should Engineer B limit involvement to a cursory review of Engineer A's AI-generated design documents, or must Engineer B provide substantive mentorship and technical oversight sufficient to close the competence gap and ensure regulatory safety compliance?

Should Engineer A conduct a full independent technical review of AI-generated design documents before sealing them, or apply standard QA protocols treating AI output as equivalent to conventional drafting tools?

Should Engineer A fully disclose AI tool usage to the client and apply heightened verification to AI-generated report and design outputs, or treat AI as an internal drafting tool requiring only standard review and no separate disclosure?

Should Engineer A seek alternative mentorship or supervisory support after losing access to a mentor, or continue the project independently given existing competence in the domain?

Should Engineer A proactively disclose AI tool usage to the client and conduct independent verification of all AI-generated design and report content before sealing and submitting deliverables, or is it sufficient to apply standard QA protocols without separate disclosure?