US-12626318-B2 - Content editing software via automatic and auditable authorship attribution

US12626318B2US 12626318 B2US12626318 B2US 12626318B2US-12626318-B2

Abstract

A content editor or a plugin thereto automatically generates authorship tokens that identify content authored by a human author or an artificial author. The authorship tokens are applied to the work while the work is being produced. Thus, subsequent review of the work can identify regions produced by a human author and other regions produced by an artificial intelligence.

Inventors

Christopher Ziolkowski
VALERIE LANCELLE
Christopher Davis

Assignees

U.S. BANK NATIONAL ASSOCIATION

Dates

Publication Date: 20260512
Application Date: 20250930

Claims (20)

1 . A method comprising: maintaining a data pipeline between a platform of a first party and a camera of a mobile device of a second party such that the data pipeline resists modification of visual content from the camera except by one or more of: an operating system of the mobile device, an application of the first party, and the platform; while maintaining the data pipeline: with the application of the first party operating on the mobile device of the second party: obtaining visual content from the camera; editing the visual content at the direction of the second party; and associating an authorship token with the visual content obtained from the camera and edited using the application, the authorship token having a predetermined form indicating that an associated portion of the visual content has non-artificial authorship rather than artificial authorship; with a first party platform of the first party: receiving, from the application of the first party, the visual content; publicly hosting, with the first party platform, the visual content of the second party; receiving, at the first party platform, a request by a third party regarding the visual content; and adjudicating the request from the third party regarding the visual content using the authorship token associated with the visual content.
2 . The method of claim 1 , wherein adjudicating the request includes determining to deny the request responsive to determining that the visual content has non-artificial authorship rather than artificial authorship based on the authorship token.
3 . The method of claim 1 , further comprising: remediating the visual content based on the request; wherein remediating the visual content includes replacing an original region of the visual content with a replacement region of artificial content produced by an artificial intelligence; and applying an artificial authorship token to the replacement region indicating that an artificial intelligence rather than a human is the author of the replacement region.
4 . The method of claim 1 , further comprising: automatically determining to continue to publicly host the visual content of the second party until adjudication is complete responsive to the visual content lacking authorship tokens indicating that the visual content has artificial authorship.
5 . The method of claim 1 , wherein the authorship token is a first authorship token; and wherein the method further comprises: applying a second authorship token to the visual content, the second authorship token having a predetermined form indicating that a second associated portion of the visual content has artificial authorship rather than non-artificial authorship.
6 . The method of claim 1 , further comprising: storing the authorship token in a block in a blockchain in association with the visual content.
7 . The method of claim 1 , with the application of the first party operating on the mobile device of the second party: obtaining additional content from outside of the data pipeline; and applying an unknown authorship token to a portion of the visual content associated with the additional content, wherein the unknown authorship token has a predetermined form indicating that the provenance of the portion of the visual content associated with the additional content is unknown.
8 . The method of claim 1 , further comprising: contemporaneous with obtaining the visual content from the camera, obtaining, with the application, additional data from one or more additional sensors; and verifying non-artificial provenance of the visual content using the additional data.
9 . A system of a first party platform for publicly hosting visual content, the system comprising: a set of one or more processors; and a set of one or more computer readable media, the set of one or more computer readable media having instructions that, when executed by one or more processors of the set of one or more processors cause the one or more processors to: receive visual content from one or more second parties; determine whether to enroll portions of the visual content into a content fingerprinting service of the first party platform based on whether the visual content includes one or more authorship tokens demonstrating that the portions of the visual content have non-artificial authorship; resist enrolling portions of the visual content into the content fingerprinting service of the first party platform responsive to the visual content having one or more authorship tokens indicating that the visual content has artificial authorship rather than human authorship; enroll visual content into the content fingerprinting service of the first party platform responsive to the visual content having one or more authorship tokens indicating that the visual content has non-artificial authorship rather than artificial authorship; publicly host the visual content; receive a request from a third party regarding subject visual content; adjudicate whether to remediate the subject visual content based on the request from the third party and based on one or more authorship tokens associated with subject visual content or based on whether the visual content is enrolled in the content fingerprinting service; remediate the subject visual content responsive to adjudicating to remediate the subject visual content.
10 . The system of claim 9 , wherein the request from the third party is to remediate the subject visual content based on the subject visual content includes an unauthorized digital replica; and wherein to remediate the subject visual content includes to replace a region of the visual content with a replacement region of artificial visual content produced by an artificial intelligence.
11 . The system of claim 9 , wherein the request from the third party is to remediate the subject visual content based on the subject visual content including infringing content; and wherein to remediate the subject visual content includes to: generate replacement content with a generative artificial intelligence; replace the infringing content with the replacement content; continuing to host the subject visual having the replacement content and not the infringing content.
12 . The system of claim 11 , wherein the instructions further cause one or more processors of the set of one or more processors to: apply an authorship token to the replacement region indicating that an artificial intelligence rather than a human is the author of the replacement region.
13 . The system of claim 9 , wherein the system further comprises: a mobile device of a second party having a camera, mobile device processors, and mobile device memory having mobile device instructions, wherein the mobile device instructions include instructions for an application of the first party that, when executed by the mobile device processors cause the mobile device processors to: maintain a data pipeline from the camera of the mobile device to the platform of the first party such that the data pipeline resists untracked editing of visual content from the camera.
14 . A method comprising: with an application of a first party operating on a mobile device of a second party having a camera: obtaining visual content from the camera of the mobile device; modifying a first region of the visual content as a result of a first manual edit to the first region of the visual content caused by a human user of the application, wherein the manual edit to the first region of the visual content includes one or more changes, insertions, or deletions to the visual content; determining that the first manual edit was caused by a human user based on either or both of: (1) the first manual edit resulting from human input received via the at least one human interface device or (2) the manual edit being provided in a fashion similar to how a human would provide the input; responsive to the determining that the first manual edit was caused by a human user, associating the first region of the visual content with a first authorship token having a form indicating that the first region has human authorship rather than artificial authorship; modifying a second region of the visual content as a result of an artificial edit to the second region of the visual content, wherein the artificial edit to the second region of the visual content includes one or more changes, insertions, or deletions to the visual content; determining that the artificial edit is from an artificial source; and responsive to the determining that the artificial edit is from an artificial source, associating the second region with a second authorship token having either (1) a form indicating that the second region has artificial authorship rather than human authorship or (2) a form indicating that the second region has mixed human authorship and artificial authorship; with a first party platform of the first party: receiving, from the application of the first party, the visual content associated with the first and second authorship tokens; publicly hosting, with the first party platform, the visual content of the second party; receiving, at the first party platform, a request from a third party regarding the visual content of the second party; and adjudicating the request from the third party regarding the visual content using one or more of the associated authorship tokens.
15 . The method of claim 14 , further comprising: enrolling the first region of visual content into a content fingerprinting service of the first party platform responsive to determining that the first region is associated with the first authorship token indicating that the first region has human rather than artificial authorship; and resisting enrolling the second region of the visual content into the content fingerprinting service of the first party platform responsive to determining that the second region is associated with the second authorship token indicating that the second region has artificial rather than human authorship.
16 . The method of claim 14 , further comprising: with the first party platform of the first party: determining that the request includes an allegation that the visual content includes an unauthorized digital replica; and remediating the visual content, wherein the remediating includes: generating replacement content for the second region of the visual content; and modifying the second region of the visual content with the replacement content.
17 . The method of claim 14 , further comprising: presenting the publicly hosted visual content via the application by the first party operating on the mobile device of the second party.
18 . The method of claim 14 , further comprising: contemporaneous with obtaining the visual content from the camera, obtaining, with the application, additional data from one or more additional sensors; and using the additional data to verify provenance of the visual content or regions thereof.
19 . The method of claim 14 , further comprising: applying at least one of the one or more authorship tokens to the visual content obtained from the camera, the one or more authorship tokens having a predetermined form indicating that the visual content from the camera has non-artificial authorship.
20 . The method of claim 14 , further comprising: with an application of the first party operating on a mobile device of the second party having a camera, maintaining a data pipeline from a camera of the mobile device to the platform of the first such that the pipeline resists editing of visual content from the camera by any editor other than the operating system of the mobile device, the application of the first party, and the platform.

Description

CROSS REFERENCE TO RELATED APPLICATIONS This application claims priority from U.S. patent application Ser. No. 19/241,881 (filed Jun. 18, 2025), which is a continuation of PCT Application No. PCT/US2025/012747 (filed Jan. 23, 2025), and is related to and claims the benefit of U.S. Provisional Patent Application Nos. 63/625,601 (filed Jan. 26, 2024), 63/638,815 (filed Apr. 25, 2024), 63/649,673 (filed May 20, 2024), 63/664,959 (filed Jun. 27, 2024), and 63/728,202 (filed Dec. 5, 2024). Each of these applications is incorporated by reference herein in their entirety for any and all purposes. BACKGROUND Traditionally, content creation software relied on human input to produce content. Thus, content could be assumed to have human authorship. Indeed, to label something as “human authored” was unnecessary because of course it was. Eventually, relatively basic or repetitive content could be generated with the help of simple software run at the creative direction of a human user (e.g., automatic creation of tables of contents based on document headers). But even then, such content was still considered human authored. Only since the arrival of large language models has artificial intelligence grown in capability sufficient to allow for the ubiquitous generation of useful human-like content with little or no input from a human author. Now, content can have one or more human authors, one or more artificial authors, or even a combination thereof. Given the quality of content produced by generative artificial intelligence, it can be difficult, if not impossible, to separate human-generated content from artificially generated content from analyzing the content alone. In addition, as new paradigms of interaction with artificial agents (e.g., which may be authors of content) develop, being able to track the provenance of content will remain useful. For instance, a human may interact with a multi-model interface or a multimodal model (e.g., GPT-4o by OPENAI) over visual, auditory, and text channels and receive an output over those same channels. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1, which is split into FIGS. 1A and 1B, illustrates an example system that can benefit from techniques described herein. FIG. 2, which is split into FIGS. 2A-H, illustrates an example method implementing techniques described herein. FIG. 3 illustrates an example text editor displaying a file having source code content that includes comments with varying styles of authorship tokens. FIG. 4 illustrates an example system implementing aspects described herein. FIG. 5 illustrates an example user interface showing a change to content. FIG. 6 illustrates an example web browser (e.g., running on a user device) rendering a web page that provides a text editor for editing content and viewing authorship information. FIG. 7, which is split into FIGS. 7A and 7B, illustrates a method for attributing authorship of content based on a conversation history. FIG. 8 illustrates an example method for using content based on its authorship. FIG. 9 illustrates an example video editor user interface having authorship attribution. FIG. 10 illustrates an example computing environment usable with techniques described herein. FIG. 11 illustrates an example machine learning framework that can benefit from or be used with techniques described herein. FIG. 12 illustrates the authorship token instructions including a method for determining authorship using embeddings. FIG. 13 illustrates an example user interface of an editor showing content. FIG. 14 illustrates an example method for storing an authorship token. FIG. 15 illustrates an example method for verifying an authorship token. FIG. 16 illustrates tracking the selection and arrangement of content. FIG. 17 illustrates an example method for determining significance based on a suggestion associated with an edit. FIG. 18 illustrates an example method for reassessing authorship tokens. FIG. 19 illustrates an example method for determining the anthropogenic status of content. FIG. 20 illustrates an example user device having a user interface showing content. FIG. 21 illustrates an example method involving a collaboration profile and visualization. FIG. 22 illustrates an example web browser running on a user device and rendering a web page that provides a content editor for editing content stored in a file. FIG. 23 illustrates a first example user interface including a chart representation of authorship information. FIG. 24 illustrates a second example user interface including a chart representation of authorship information. FIG. 25 illustrates a third example user interface including a chart representation of authorship information. FIG. 26 illustrates an example user interface for visualizing a relative change in authorship of regions of a file over the course of an editing session. FIG. 27 illustrates an example method for modifying artificial involvement based on a contribution history. FIG. 28 illustrates an example metho