Substantial Compositional Contribution Requires Attribution

Source: jason@calacanis.com. "Disgraziad: The WSJ’s Millie Vanilli Editorial Page." August 29, 2026. jcalfromallin.substack.com

The Gist

If someone else did enough of the writing, you already put them in the byline. A model that drafts the piece is that someone else. Say so.

Conclusion

Undisclosed AI composition breaches an attribution norm that publishing has long recognized, and the fix is a credit line, not a prohibition.

Premises

1. Publishing under a byline asserts that the named person originated both the reasoning and its expression.
2. Where a second party contributes substantially to the expression, publishing conventions already require credit. The "with" byline exists for this case, and concealing such an arrangement is understood as deceiving the reader rather than as normal practice.
3. That convention is indifferent to whether the contributor is human. It attaches to the fact of substantial compositional contribution, not to the contributor's nature.
4. A model that drafts or substantially rewrites a piece meets the same threshold that triggers the convention for a human collaborator.
5. Undisclosed documentary re-enactment is the correct parallel: the represented content may be accurate, but the audience is misled about the epistemic status of what it is receiving.
6. The remedy is already available and requires no new ethics, only application: name the model in the byline.

Assumptions

• Primary sources: Jason Calacanis, All-In Podcast clip posted by @ThisWeeknAI, 29 Aug 2026 (https://x.com/ThisWeeknAI/status/2093770140730957969); and the companion essay "Disgraziad: The WSJ's Millie Vanilli Editorial Page," JCal from All In, 29 Aug 2026 (https://jcalfromallin.substack.com/p/disgraziad-the-wsjs-millie-vanilli). All-In episode chapter "AI writing controversy" at 1:01:33 (https://podcasts.apple.com/us/podcast/all-in-with-chamath-jason-sacks-friedberg/id1502871393). This reconstruction is of Calacanis's published case, not of Druckenmiller's or the WSJ's.
• Steelman of Calacanis's essay-informed position. "More than their share" is left as his threshold, not operationalized here.
• The essay's permitted uses (research, brainstorming, proofing, shortening without changing anything important) are treated as below the attribution threshold. That line is asserted, not derived.
• A characterization of training data as stolen, and the sample byline naming a Chinese open-source model, are not premises.

Analysis

Overall strength: Moderate. Argument type: Deductive.

Premise Strength

• Publishing under a byline asserts that the named person originated both the reasoning and its expression. (Moderate) — Plausible as an idealized reader expectation, but treats a contested and unevenly practiced convention as settled fact; heavy editorial involvement, house style changes, and PR-drafted statements already complicate this claim in existing practice.
• Where a second party contributes substantially to the expression, publishing conventions already require credit... (Weak) — Rests on a single testimonial source generalizing to an industry-wide norm; real-world ghostwriting practice, especially in opinion journalism, is inconsistently disclosed, which undercuts the claim that concealment is uniformly 'understood as deceiving the reader' rather than tolerated practice.
• That convention is indifferent to whether the contributor is human. (Weak) — This is the argument's central, load-bearing generalization, explicitly asserted rather than derived (per A3). Human-collaborator norms may be entangled with agency, consent, and compensation considerations that do not straightforwardly extend to a non-agentive tool, making the extension a genuine point of contest rather than an established feature of the prior convention.
• A model that drafts or substantially rewrites a piece meets the same threshold that triggers the convention for a human collaborator. (Weak) — Depends entirely on the undefined 'substantial contribution' threshold (A2), making the claim unfalsifiable as stated and unable to be checked against any actual case, including the one motivating the argument.
• Undisclosed documentary re-enactment is the correct parallel... (Moderate) — Effective as illustrative rhetoric and captures the 'misled about epistemic status' point, but the analogy has an unaddressed disanalogy: re-enactments misrepresent facts-in-the-world, while AI drafting misrepresents production process, arguably a different order of harm.
• The remedy is already available and requires no new ethics, only application: name the model in the byline. (Moderate) — Follows reasonably if prior premises are granted, but understates real implementation costs — no verification or enforcement mechanism exists, and the norm relies on self-disclosure by parties with incentive to conceal, exactly the behavior that precipitated the controversy.

Potential Fallacies

• Begging the question (petitio principii) (P3) — The claim that the attribution convention is indifferent to the contributor's nature is presented as an established feature of an existing norm, when it is actually the contested extension the argument needs to prove. The convention was observed and justified in human-to-human contexts; treating its scope as already settled for non-human contributors assumes what must be shown.
• Hasty generalization from limited testimony (P2) — The claim that publishing conventions 'already require' credit for substantial contribution is generalized from a single commentator's characterization rather than from editorial style guides, contracts, or documented practice. In reality, ghostwriting disclosure is inconsistently applied, especially in the exact genre (public-figure op-eds) under discussion, which weakens the premise's claim to describe a settled, enforced norm.
• Weak or false analogy (P5) — Undisclosed documentary re-enactment misrepresents whether depicted events actually occurred; undisclosed AI drafting misrepresents only the process by which existing claims were expressed, not necessarily the truth or ownership of those claims. Treating these as 'the correct parallel' imports a moral weight (deception about reality) that may not transfer cleanly to deception about composition process.
• Vague or unoperationalized standard (P4, supported by A2/A3) — The pivotal threshold — 'substantial contribution' or 'more than their share' — is explicitly left undefined. A rule this indeterminate cannot be consistently applied to real cases, inviting selective enforcement, threshold-gaming (redefining drafting as 'proofing' or 'shortening'), and disputes that replicate the very controversy the norm is meant to resolve.
• Unstated proportionality premise (enthymeme) (P6 / Conclusion) — Even granting that a norm violation occurred, the conclusion that the appropriate remedy is disclosure rather than prohibition requires an additional premise — that remedy-type should match violation-type — which is asserted in P6 rather than argued for, and forecloses the live alternative of stricter limits without addressing it.

Counterarguments

• Premise P3 (High impact) — The human ghostwriting convention may be justified by concerns specific to human agents — compensation, consent, reputational stake, and accountability — none of which apply to a non-agentive AI system. If so, the convention is not indifferent to contributor nature but actually presupposes it, undermining the generalization the argument needs.
• Premise P4 / Conclusion (High impact) — An accountability-based model of authorship — under which what matters is whether the named author endorses, understands, and can defend the substantive claims, not who physically produced the expression — dissolves the need for an operationalized 'substantial contribution' threshold entirely and better matches how editing, translation, and tool-assistance have long been tolerated without disclosure.
• Premise P2 (High impact) — Ghostwriting disclosure is demonstrably inconsistent in practice — political speeches, executive-bylined statements, and many op-eds are routinely published without 'with' credit and are not treated as scandalous — which weakens the claim that an established, binding convention already exists for the human case that AI can simply be folded into.
• Premise P5 (Medium impact) — Documentary re-enactment and Milli Vanilli-style fraud involve deception about facts-in-the-world (whether events occurred as depicted, whether a performance was live); undisclosed AI drafting need not misrepresent any fact about the world, only the process of composition, making the analogy's imported moral weight potentially disproportionate.
• Premise P6 / Conclusion (Medium impact) — Declaring the remedy 'requires no new ethics, only application' ignores that there is no mechanism to verify how much AI contributed to a given piece; the norm relies entirely on voluntary self-report by the same actors who had incentive to conceal, making it largely symbolic against the cases it targets.
• Conclusion (High impact) — Extending the 'substantial contribution triggers disclosure' rule consistently would require 'with' credits for the vast majority of heavily edited op-eds, ghostwritten speeches, and advisor-shaped academic papers — an implication neither addressed nor plausible, suggesting the rule as stated is either radically underapplied today or incoherent when generalized.

Suggested Improvements

• Operationalize the central threshold — Replace 'substantial contribution' and 'more than their share' with a measurable criterion (e.g., proportion of final text originating from AI output, or a defined functional test distinguishing drafting from editing). Without this, the argument's central premise (P4) cannot be applied consistently or falsified, which is the most frequently identified critical weakness across independent scrutiny.
• Substantiate the empirical claim in P2 — Cite actual editorial style guides, publishing contracts, or documented industry practice showing the 'with' credit convention is consistently enforced, rather than relying on a single commentator's characterization. Current practice suggests ghostwriting disclosure is inconsistent, especially in opinion journalism, which undermines the premise's descriptive accuracy.
• Engage the accountability-based counter-model directly — Address the strongest opposing view — that what matters is whether the named author endorses and can defend the content, not who authored its expression — rather than only contrasting AI with word processors and spell-checkers. This counter-model is harder to dismiss than the tool-analogy objection already addressed in the source, and sidesteps the unoperationalized threshold problem entirely.
• Address enforcement and verification — Supplement the disclosure remedy with a discussion of detection mechanisms, incentive structures, or institutional accountability, rather than relying solely on voluntary self-report. Without enforcement, the proposed remedy is vulnerable to the same concealment incentives that produced the controversy prompting the argument.
• Separate the norm-violation claim from the remedy claim — Explicitly argue for the proportionality premise connecting 'a disclosure norm was breached' to 'therefore disclosure, not prohibition, is the correct remedy,' rather than asserting it. This step is currently an enthymeme; making it explicit would let readers assess it rather than accept it by default.

Scenario Tests

• A human author uses AI to generate a full first draft but then reviews every claim, edits extensively for voice, and can independently defend all reasoning and assertions in the piece. (Challenges) — Suggests the accountability-based model may be more relevant than pure origination — if the named author genuinely stands behind the content, the deception concern in P1/P5 may be substantially reduced regardless of drafting mechanics.
• A political or corporate op-ed is entirely ghostwritten by a human professional without any 'with' credit, following common industry practice. (Challenges) — Directly undermines P2's claim that the convention is already consistently enforced for human collaborators, weakening the analogical bridge to AI.
• An AI is used only to suggest phrasing sentence-by-sentence, with the human directing every substantive point and idea. (Neutral) — Exposes the threshold problem directly: this case sits ambiguously between 'editing' and 'drafting' under the argument's own unoperationalized standard, and could be classified either way depending on motivated interpretation.
• A documentary broadcasts an accurate but undisclosed staged re-enactment of a historical event. (Supports) — Validates the core intuition behind P5 — that accuracy of content doesn't preclude audience deception about epistemic status — lending some support to the general principle, even if the disanalogy to AI-drafting's process-only deception remains a caveat.
• Reader surveys show no significant change in trust, perceived quality, or feelings of deception when AI involvement is disclosed versus undisclosed for opinion pieces. (Challenges) — Would undercut the empirical assumption embedded in P1 and P5 that readers are meaningfully harmed or misled by non-disclosure, suggesting the norm may be more aspirational than reflective of actual reader concern.

Coherence & Relevance

The argument is internally coherent as a syllogism: granting P3 and P4 yields the norm-violation portion of the conclusion validly. Its overall soundness, however, depends heavily on premises that are conceded within the argument's own assumptions to be asserted rather than derived (P3's nature-indifference, P4's threshold), and on an empirical premise (P2) whose accuracy is contestable given inconsistent real-world enforcement of ghostwriting disclosure. The remedy portion of the conclusion additionally requires an unstated proportionality principle. The result is a well-structured, rhetorically effective argument whose logical skeleton is sound but whose load-bearing joints remain underdetermined and open to substantive challenge.

• Publishing under a byline asserts that the named person originated both the reasoning and its expression. (Moderate) — Establishes the baseline reader expectation the argument needs, but idealizes byline conventions beyond how they function in practice across genres (news vs. op-ed vs. memoir).
• Where a second party contributes substantially to the expression, publishing conventions already require credit... (Strong) — Directly supplies the precedent the argument extends to AI, but its empirical accuracy is contested; the convention's consistency of enforcement is asserted rather than documented.
• That convention is indifferent to whether the contributor is human. (Strong) — This is the pivotal bridge premise; it is essential to the argument's validity but is stipulated rather than independently argued for, making it the single most important gap in the chain.
• A model that drafts or substantially rewrites a piece meets the same threshold that triggers the convention for a human collaborator. (Strong) — Directly required for the syllogism to reach its conclusion, but the threshold it invokes is explicitly left unoperationalized, so its truth cannot be checked against any concrete case.
• Undisclosed documentary re-enactment is the correct parallel... (Moderate) — Functions as illustrative/analogical support rather than a strictly necessary link in the deductive chain; strengthens intuitive plausibility but carries unaddressed disanalogies.
• The remedy is already available and requires no new ethics, only application: name the model in the byline. (Moderate) — Follows from the prior premises if granted, but the transition from 'norm violated' to 'this specific remedy, not prohibition' requires an additional unstated proportionality premise not supplied in the argument.

View this argument on LogicFirst.ai