Deepfakes at Face Value: Image and Authority
Pith reviewed 2026-05-10 14:29 UTC · model grok-4.3
The pith
Deepfakes can be wrongful by usurping a person's authority over their own image and identity, even when no harm occurs.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The paper's central claim is that deepfakes are wrong when they usurp our authority to determine the provenance of our own agency by exploiting our biometric features as a generative resource. In particular, we have a specific right against the algorithmic conscription of our identity. This right is violated even in the absence of harm, and the paper distinguishes it from permissible forms of appropriation such as artistic depiction.
What carries the argument
The right against the algorithmic conscription of identity, which protects a person's authority over the permissible uses of their image and the governance of their identity.
If this is right
- Deepfakes remain wrongful even when they cause no harm or violate no other interests, solely because they bypass authority over identity.
- Artistic depictions of a person do not count as algorithmic conscription and therefore do not violate the right.
- The wrong attaches to the act of turning biometric features into generative resources without the person's determination of provenance.
- Protection against this conscription would limit certain uses of personal data in synthetic media creation.
Where Pith is reading between the lines
- The framework could apply to other forms of AI-generated content that simulate personal traits, such as voice or motion.
- It points toward consent standards focused specifically on identity governance rather than only on data ownership.
- Public acceptance of non-harmful deepfakes of public figures might serve as a test of whether the authority interest is widely recognized.
Load-bearing premise
Individuals have a legitimate interest in authority over the permissible uses of their image and the governance of their identity that exists independently of harm or other non-normative interests, and that algorithmic simulation violates this interest in a way artistic depiction does not.
What would settle it
A concrete case of a non-consensual deepfake that produces no harm, no deception, and no other setback, yet the subject still experiences or asserts a violation of their authority over how their biometric likeness is used.
read the original abstract
Deepfakes are synthetic media that superimpose or generate someone's likeness on to pre-existing sound, images, or videos using deep learning methods. Existing accounts of the wrongs involved in creating and distributing deepfakes focus on the harms they cause or the non-normative interests they violate. However, these approaches do not explain how deepfakes can be wrongful even when they cause no harm or set back any other non-normative interest. To address this issue, this paper identifies a neglected reason why deepfakes are wrong: they can subvert our legitimate interests in having authority over the permissible uses of our image and the governance of our identity. We argue that deepfakes are wrong when they usurp our authority to determine the provenance of our own agency by exploiting our biometric features as a generative resource. In particular, we have a specific right against the algorithmic conscription of our identity. We refine the scope of this interest by distinguishing between permissible forms of appropriation, such as artistic depiction, from wrongful algorithmic simulation.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper claims that harm-based or non-normative-interest accounts fail to explain the wrongness of deepfakes in no-harm cases. It identifies a distinct wrong in the subversion of legitimate interests in authority over the permissible uses of one's image and the governance of one's identity: deepfakes usurp authority to determine the provenance of one's own agency by exploiting biometric features as a generative resource for algorithmic simulation. The paper asserts a specific right against such algorithmic conscription and refines its scope by distinguishing this from permissible appropriation, such as artistic depiction.
Significance. If the central distinction and authority-based right hold, the paper supplies a non-harm-dependent normative foundation for regulating deepfakes and similar AI-generated likenesses. This could inform ethical guidelines, legal doctrines on digital identity, and policy on biometric data use in generative AI, extending beyond existing privacy or defamation frameworks.
major comments (2)
- [Refining the Scope of the Interest] The section refining the scope of the interest (distinguishing artistic depiction from algorithmic simulation): the claim that only algorithmic conscription of biometric features usurps authority over provenance requires explicit counterexample analysis or criteria to show why non-algorithmic uses (e.g., paintings or photographs) do not trigger the same violation; without this, the distinction remains under-supported for the central normative conclusion.
- [The argument identifying the neglected reason] The development of the authority interest independent of harm: the manuscript grounds the right in stated legitimate interests but does not address how this interest would be weighed against competing claims such as free expression or public-domain uses of likeness; this is load-bearing because the no-harm case relies on the interest being sufficiently robust and non-reducible.
minor comments (2)
- [Abstract] The abstract is clear but could explicitly flag the paper's contribution as a rights-based rather than harm-based account in the opening sentence.
- [Introduction] Notation for key terms such as 'provenance of agency' and 'algorithmic conscription' is introduced without a dedicated definitions subsection; a short glossary or consistent italicization would aid readability.
Simulated Author's Rebuttal
We thank the referee for their constructive and detailed comments, which highlight important areas where the manuscript's arguments can be clarified and strengthened. We respond to each major comment below and outline the revisions we will undertake.
read point-by-point responses
-
Referee: [Refining the Scope of the Interest] The section refining the scope of the interest (distinguishing artistic depiction from algorithmic simulation): the claim that only algorithmic conscription of biometric features usurps authority over provenance requires explicit counterexample analysis or criteria to show why non-algorithmic uses (e.g., paintings or photographs) do not trigger the same violation; without this, the distinction remains under-supported for the central normative conclusion.
Authors: We accept that the distinction between permissible artistic appropriation and wrongful algorithmic simulation requires more explicit support to sustain the central normative claim. The manuscript grounds the distinction in the automated conscription of biometric features as a generative resource for simulation, which bypasses individual authority over provenance in a way that manual artistic rendering does not. To address the concern, we will revise the relevant section to include explicit criteria (such as the presence of algorithmic feature extraction and automated replication versus human interpretive choice) and provide counterexample analysis. This will cover traditional paintings, hand-drawn portraits, and conventional photography, explaining why these do not usurp authority over the provenance of one's agency in the same manner, as they lack the direct, non-consensual algorithmic exploitation of biometric data. These additions will make the scope of the interest clearer without altering the paper's core thesis. revision: yes
-
Referee: [The argument identifying the neglected reason] The development of the authority interest independent of harm: the manuscript grounds the right in stated legitimate interests but does not address how this interest would be weighed against competing claims such as free expression or public-domain uses of likeness; this is load-bearing because the no-harm case relies on the interest being sufficiently robust and non-reducible.
Authors: The manuscript's primary contribution is to establish the authority interest as a distinct, non-reducible normative reason for the wrongness of deepfakes that operates even in the absence of harm or other setbacks to non-normative interests. We do not present this right as absolute. To better demonstrate its robustness, we will add a brief discussion clarifying that the interest is pro tanto and must be weighed against competing considerations such as free expression (e.g., in parody or satire) and certain public-domain or historical uses of likeness. This addition will note that such balancing does not reduce the interest to harm-based or other accounts but preserves its independent force in no-harm cases. We view this as a clarification rather than a fundamental expansion of the argument. revision: partial
Circularity Check
No significant circularity
full rationale
The paper advances a normative ethical argument identifying a right to authority over the provenance of one's agency and permissible uses of one's image, violated specifically by algorithmic conscription in deepfakes. This interest is stated independently of harm-based accounts and refined internally by distinguishing artistic depiction from wrongful simulation. No equations, fitted parameters, self-citations, or uniqueness theorems are invoked in the provided text. The derivation does not reduce the target right to a self-referential definition or prior result by construction; the central claim retains independent content grounded in the stated interests.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption Individuals have legitimate interests in having authority over the permissible uses of their image and the governance of their identity.
- ad hoc to paper Algorithmic simulation of biometric features constitutes a distinct wrongful appropriation that usurps authority, unlike permissible forms such as artistic depiction.
Reference graph
Works this paper leans on
-
[1]
Introduction Deepfakes are synthetic media that superimpose or generate someone’s likeness onto pre-existing sound, images, or videos using deep learning (DL) methods.1 Previously, the production of fake content was accessible only to photo-edited and computer-generated imagery experts. But with recent advances in DL and the Internet, manipulated media is...
work page 2017
-
[2]
Deepfakes and their harms A common view is that it is wrong to create and distribute deepfakes because of the harms that this can cause. The Harm-Based Account. Deepfakes are morally wrong when they cause or constitute harm to someone in some kind of way. The Harm-Based Account is appealing in its simplicity, although this characterisation leaves open sev...
work page 2021
-
[3]
The problem of bare wronging While many deepfakes are harmful for exactly the reasons outlined above, these harms do not exhaust the reasons why deepfakes can be morally wrong. The Harm-Based Account assumes that the set of wrongful acts is entirely contained within the set of harmful acts. However, there is a class of cases, so-called harmless or bare wr...
work page 2020
-
[4]
I claim, the moral intuition of most people is that B is doing something wrong
Öhman writes: “I claim, the moral intuition of most people is that B is doing something wrong” (2020: 134). We take it that most people would agree with Öhman that it is wrong for B to create the deepfake of A in this manner, even if it is impossible to distribute it or for anyone else, including the victim, to come to know of its existence. Nevertheless,...
work page 2020
-
[5]
Gender oppression, control, and other considerations How can we explain the moral wrongness in the mere creation of a pornographic deepfake like Pornographic Deepfake, if it does not cause any identifiable harms? A natural place to start is with a suggestion made by Öhman himself. He argues that the intuition that B is doing something wrong by creating a ...
work page 1998
-
[6]
present them in ways that disregard their will and go against their sense of self
We do not deny the force of this explanation. Öhman is likely correct that the vast majority of actual deepfake pornography constitutes a specific form of gendered violence that functions to uphold patriarchal oppression. However, while this account explains the severity and systemic function of the wrong in many cases, it does not isolate the constitutiv...
work page 2023
-
[7]
In particular, we defend the following hypothesis: The Authority Interest View
The Authority Interest View and The Right Against Algorithmic Conscription In this section, we argue that the creation of the deepfakes in Pornographic Deepfake and Pornstar’ s Deepfake are moral wrongs against their subjects because they undermine the subjects’ interests in having authority over who does what with their image. In particular, we defend th...
work page 2001
-
[8]
explains cases of bare wrongings in terms of our normative interests: interests which we have in controlling certain features of the normative landscape, such as the obligations and commitments we have and make to one another, our own and other’s rights and responsibilities, or the information that can permissibly be made available to each other. We have ...
work page 2012
-
[9]
New Media & Society 23(7): 2072–2098
Anticipating and addressing the ethical implications of deepfakes in the context of elections. New Media & Society 23(7): 2072–2098. Dines, G., B. Jensen, and A. Russo
work page 2072
-
[10]
The Verge https://www.theverge.com/2024/8/19/24223589/trump-ai-generated-swift-harris-social-media
Donald Trump posts a fake AI-generated Taylor Swift endorsement. The Verge https://www.theverge.com/2024/8/19/24223589/trump-ai-generated-swift-harris-social-media. Fletcher, J
work page 2024
-
[11]
https://www.theguardian.com/technology/artificialintelligenceai/2022/mar/19/all Mogensen, K
Deepfakes v pre-bunking: Is Russia losing the infowar? The Guardian. https://www.theguardian.com/technology/artificialintelligenceai/2022/mar/19/all Mogensen, K. Æ
work page 2022
-
[12]
Scripps-Howard Broadcasting Co., 433 U.S
Zacchini v. Scripps-Howard Broadcasting Co., 433 U.S. 562 (1977) Statements & Declarations Funding: The author declares that no funds, grants, or other support were received during the preparation of this manuscript. Competing Interests: The author has no relevant financial or non-financial interests to disclose
work page 1977
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.