DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Continued Examination Under 37 CFR 1.114
2. A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on June 10th 2025 has been entered.
Response to Amendment
3. The Amendment filed on June 10th 2025 has been entered. Claims 1, 6, 11 and 16 have been amended, claims 2 – 5, 7, 8, 12 – 15, 17, 18 and 21 are cancelled, claims 22 – 33 are newly added with claims 1, 6, 9 – 11, 16, 19, 20 and 22 - 33 are pending in the application.
Response to Arguments
35 U.S.C. §102
4. Applicant's arguments, see Remarks pp. 7 -9, filed June 22nd 2026, with
respect to the rejections of claims 1,6, 9 – 11, 16, 19, 20, 24, 26, 27, 30, 32 and 33 under 35 U.S.C. §102 have been fully considered and they are persuasive.
The crux of applicant’s arguments is that the antecedent priority of the Horton reference as prior art is improper.
Examiner respectfully agrees and withdraws the rejection
Upon further consideration relevant prior art has been found to examine applicant’s claimed invention and are made in view of Tianxiang et al. (Korean Patent Publication Number 2023-0088381), hereinafter referred to as Tianxiang, Beauchesne et al., (United States Patent Publication Number 2023/0143574) hereinafter Beauchesne and Jana Dittman et al., (Using Cyrptographic and Watermarking Algorithms) hereinafter Dittman
Claim Rejections – 35 U.S.C. §103
5. The following is a quotation of 35 U.S.C. 103 which forms the basis for all
obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
6. The factual inquiries set forth in Graham v John Deere Co., 383 U.S. 1, 148 USPQ
459 (1966), that are applied for establishing a background for determining obviousness
under 35 U.S.C. 103 are summarized as follows:
a. Determining the scope and contents of the prior art
b. Ascertaining the differences between the prior art and the claims at issue
c. Resolving the level of ordinary skill in the pertinent art
d. Considering objective evidence present in the application indicating
obviousness or nonobviousness
Claims 1, 6, 9 – 11, 16, 19, 20, 24, 26, 30 and 32 are rejected under 35 U.S.C. 103 as being unpatentable over Beauchesne et al., (United States Patent Publication Number 2023/0143574) hereinafter Beauchesne in view of Sarkissian et al., (United States Patent Publication Number 2024/0311659)
Regarding claim 1 Beauchesne teaches a system (ABS., system) (Figs. 1, 10 and 11 systems [0008], [0020], [0022]) for identifying artificial intelligence (Al) generated content, comprising: a plurality of computing devices (cluster of computing devices [0122]) each comprising at least a processor, (at least one processor [0116]) a memory, (including the system memory 1230 such as read-only memory (ROM) 1240 and random access memory (RAM) 1250 [0122]) and a network interface; (Fig. 12 communication interface [0125]) wherein a plurality of programming instructions stored in one or more of the memories (computer-readable storage medium to store instructions [0123])and operating on one or more of the processors (at least one processor [0116]) of the plurality of computing devices (cluster of computing devices [0122])causes the plurality of computing devices to: (cluster of computing devices [0122]) receive input content (once the system receives the content [0027]) comprising a plurality of fingerprints, (The fingerprints are (for example, with video content) abstract, non-reversible representations of scenes, segments, clips, and/or subportions
of the media content. [0025]) wherein each fingerprint comprises at least one hash value; (ABS., These fingerprints are mathematical vectors generated using one or more techniques, such as perceptual hashes coupled with machine
learning) (Fig. 1, (116) video hash, (118) audio hash and (128) keyframe hash) [0083], [0084]) compare the hash values of the input content (the content [0027]) against a registry of content known (The system 100 then respectfully compares 130, 132, 134 the video hash 116, the audio hash 118, and the keyframe hash 128 to fingerprints of known content 120 [0084]) the registry (fingerprints of known content 120 [0084]) such as “registry” comprising a plurality of registered hash values organized into content groups; (known (audio) fingerprints 120; known (video) fingerprints 120 and known (keyframe) fingerprints 120 [0084]) such as “content groups” and for each identified registered content group (any one of (known (audio) fingerprints 120; known (video) fingerprints 120 and known (keyframe) fingerprints 120 [0084]) such as “content groups” having at least one matching element ( respective matching sections [0101])with the input content: (the content [0027]) determine one or more portions ( portions of the audio 802, video 804, and keyframe 808 [0101]) of the input content (the content [0027]) which match (As illustrated, there are portions of the audio 802, video 804, and keyframe 808 comparisons which are matching 812 between the uploaded content and the known content. [0101]) the registered content group (known (audio) fingerprints 120; known (video) fingerprints 120 and known (keyframe) fingerprints 120 [0084]) such as “content groups” by comparing hash values of the input content portions against hash values(Fig. 6A comparison results for audio hashes (match : 88.04% [0013], [0096]) stored in the registered content group: (known (audio) fingerprints 120; known (video) fingerprints 120 and known (keyframe) fingerprints 120 [0084]) such as “content groups” and determine a confidence level (This large quantity of fingerprints per item or sub-items of content increases the confidence level and accuracy of the decisions taken by matching engines in the later stages of the scanning p [0025]) based on statistical analysis of multiple similarity metrics (Similar faces can be clustered based on a similarity threshold determined for each embedding model and then, in each cluster, the best face can be selected based on a "matchability" score. [0058]) indicating a likelihood(likelihood of a match [0037]) that the determined portion ( portions of the audio 802, video 804, and keyframe 808 [0101]) of the input content (the content [0027]) associated with the registered content group; (known (audio) fingerprints 120; known (video) fingerprints 120 and known (keyframe) fingerprints 120 [0084]) such as “content groups” and aggregate results (results are aggregated [0049]) from multiple registered content groups (known (audio) fingerprints 120; known (video) fingerprints 120 and known (keyframe) fingerprints 120 [0084]) such as “content groups” to indicate a proportion ( portions of the audio 802, video 804, and keyframe 808 [0101]) of the input content(the content [0027]) based at least on the determined portions ( portions of the audio 802, video 804, and keyframe 808 [0101]) and confidence levels (any content having a level of similarity above a threshold amount [0075]) from each registered content group. (known (audio) fingerprints 120; known (video) fingerprints 120 and known (keyframe) fingerprints 120 [0084]) such as “content groups”
Beauchesne does not fully disclose to have been previously generated by one or more generative AI services, was generated by a generative AI service, that was likely generated by any generative AI service
Sarkissian teaches to have been previously generated by one or more generative AI services, (content generating source systems
175, such as generative artificial intelligence systems [0378]) was generated by a generative AI service, (content generating source systems 175, such as generative artificial intelligence systems [0378])that was likely generated by any generative AI service(content generating source systems 175, such as generative artificial intelligence systems [0378])
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Beauchesne to incorporate the teachings of Sarkissian wherein to have been previously generated by one or more generative AI services, was generated by a generative AI service, that was likely generated by any generative AI service. By doing so can provide one or more advantages to prior approaches, including the tracking, monitoring, authenticating, source-validating, managing, and governing of generated output and content, such as, for example, artificial intelligence generated output. Sarkissian [0016]
Claim 11 corresponds to claim 1 and is rejected accordingly
Regarding claim 6 Beauchesne in view of Sarkissian the system of claim 1,
Beauchesne as modified further teaches wherein the registry further comprises content created by human users and protected content including copyrighted or licensed material (where the known content is copyrighted and/or
which was previously identified as prohibited content (such as child pornography, non-consensual media, etc.). Prohibited content, as defined herein, can include any content which the owner or up loader of the media file is not legally
allowed to share, including copyrighted media, video or images of underage sexual content, video or images which were obtained without consent, etc. [0024])
Claim 16 corresponds to claim 6 and is rejected accordingly
Regarding claim 9 Beauchesne in view of Sarkissian the system of claim 1, Beauchesne as modified further teaches wherein the computing devices (cluster of computing devices [0122])further: produce a report based on analysis (These metrics and overlaps can be converted into a summary response, where all matching content is ranked according to the amount of time (seconds) of overlap. In other configurations, the ranking could be based on percentages of similarity or other similarity metrics [0075]) of the input content, (the content [0027])wherein the report ( a summary response, [0075])comprises at least the one or more identified registered content groups; ( The most likely fingerprints can be grouped by content (video, image, or audio), and an overlap can be calculated between the fingerprints results and the query, preferably in seconds of overlap though other metrics are possible. For example, based on the overlap of video and audio fingerprints, the system can calculate a 300 seconds (five minutes) of overlap between the two pieces of content. [0075]) and for each identified registered content group, (known (audio) fingerprints 120; known (video) fingerprints 120 and known (keyframe) fingerprints 120 [0084]) such as “content groups” include the determined proportion (The UI can also show a percentage amount identifying how much of the known content matches the content under review, [0102]) SEE Fig. 9A and the determined confidence level (and provide the user options such as "Copy Response," "Confirm Match," "Dismiss Match," etc. [0102])
Claim 19 corresponds to claim 9 and is rejected accordingly
Regarding claim 10 Beauchesne in view of Sarkissian the system of claim 1, Beauchesne as modified further teaches wherein the input content (the content [0027]) is multimedia content (ABS., image, audio, and/or video aspects of the media.) (video 508 [0095])
Claim 20 corresponds to claim 10 and is rejected accordingly
Regarding claim 23 Beauchesne in view of Sarkissian the system of claim 1,
Beauchesne as modified further teaches comprising a generative artificial intelligence content verification exchange (GenAI CVX), (a convolutional neural network (CNN) called Deep Perceptual Hasher (DPH) [0043]) such as “a generative artificial intelligence content verification exchange (GenAI CVX)” wherein the GenAI CVX (a convolutional neural network (CNN) called Deep Perceptual Hasher (DPH) [0043]) such as “a generative artificial intelligence content verification exchange (GenAI CVX)”is configured to: receive the input content; (receiving the new movie 208, [0087]) maintain the registry of content (Each of the previously registered movies 202 has been hashed/fingerprinted/converted into an embedding … and those fingerprints are stored in a ANN DB 206, referred to as a ANN DB. [0086]) known to have been previously generated (previously registered/fingerprinted movies 202. Each of the previously registered movies 202 has been hashed/fingerprinted/converted into an embedding [0086]) by one or more generative Al services; (several embedding models [0056]) such as “one or more generative Al services” and perform comparison and analysis operations (a movie 208 is being compared to previously registered/fingerprinted movies 202. [0086]) to determine content uniqueness (The fingerprint 210 is compared to the fingerprints 204 of the previously known movies 202 and content. If the comparison reveals no clear match, the new movie 208 can proceed to candidate verification 212, where additional information about the new movie 208 may be required [0087]) and whether the input content (the new movie 208, [0087]) was generated by(In addition, unless the new movie 208 fingerprint 210 is an exact match to a previously known fingerprint 204 stored in the ANN DB 206, the new fingerprint 210 can be stored in the ANN DB 206 for comparison against future content uploads. [0087]) a generative Al service, (several embedding models [0056]) such as “one or more generative Al services” independent of watermark detection (The fingerprint 210 is compared to the fingerprints 204 of the previously known movies 202 and content [0087])
Claim 29 corresponds to claim 23 and is rejected accordingly
Regarding claim 24 Beauchesne in view of Sarkissian the system of claim 1, Beauchesne as modified further teaches wherein hash values are derived from both content data and metadata (The UI can also display metadata
(title, name of the uploading user, identifying tokens, upload date, etc.), and the matching portions/segments of the different comparisons 922 [0102]) associated with the content, (the content [0027]) the metadata (The UI can also display metadata [0102]) comprising at least one of: a device identifier, an IP address, a geographic location, or a timestamp, (upload date, [0102])and wherein a plurality of hash values (the video hash 116, the audio hash 118, and the keyframe hash 128 [0084]) representing a single piece of content (a video [0093]) are processed by a machine learning model (using deep machine learning [0093])to generate a unique content identifier ( then use a Deep Perceptual Hash 406 algorithm as described above to create a 128 bit float embedding (fingerprint) 408. [0093])
Claim 30 corresponds to claim 24 and is rejected accordingly
Regarding claim 26 Beauchesne in view of Sarkissian the system of claim 1, Beauchesne as modified further teaches wherein the comparison and analysis operations are selectively triggered by at least one of: a time-based trigger occurring at predetermined intervals; (The system 100 then respectfully compares 130, 132, 134 the video hash 116, the audio hash 118, and the
keyframe hash 128 to fingerprints of known content 120. In some configurations, these comparisons 130, 132, 134 can occur simultaneously, whereas in other configurations the comparisons 130, 132, 134 can occur sequentially. For
example, the known (audio) fingerprints 120 can be compared to the newly generated audio fingerprint 118 at the same time as the known (video) fingerprints 120 are compared to the newly generated video fingerprint 116. Alternatively,
the known (keyframe) fingerprints 120 may not be compared to the newly generated keyframe fingerprint 128 until after the comparison 130 of known (video) fingerprints
120 to the newly generated video fingerprint 116 is complete. [0084]) an event-based trigger responsive to submission of the input content; (In some configurations this sequential process of fingerprint comparisons can be dependent on a threshold level of similarity being determined by a first fingerprint
comparison. For example, the keyframe comparison 134 may not occur unless the video comparison 130 and/or the audio comparison 132 is first completed and indicates that potentially prohibited content has been detected [0084]) or a manual trigger for batch processing of stored content.
Claim 32 corresponds to claim 26 and is rejected accordingly
Claims 22, 25, 28 and 31 are rejected under 35 U.S.C. 103 as being unpatentable over Beauchesne et al., (United States Patent Publication Number 2023/0143574) hereinafter Beauchesne in view of Sarkissian et al., (United States Patent Publication Number 2024/0311659) and in further view of Tianxiang et al. (Korean Patent Publication Number KR20230088381), hereinafter referred to as Tianxiang .
Regarding claim 22 Beauchesne in view of Sarkissian teaches the system of claim 1,
Beauchesne as modified does not fully disclose wherein the input content comprises deepfake content, and the system is configured to identify portions of the deepfake content that match registered AI- generated content without relying on watermark detection
Tianxiang teaches wherein the input content (input audiovisual data Page 8) such as “input content” comprises deepfake content, (the likelihood that it contains a deepfake of an interlocutor Page 9) and the system (system 100 Page 11) is configured to identify portions of the deepfake content that match (For interlocutor deepfake detection, a machine learning architecture is used to scan an inbound audiovisual data sample indicating the likelihood that it contains a deepfake of an interlocutor based on the similarity between one or more preconfigured or registered spoofprints and the interlocutor in the inbound spoofprint Page 9) registered AI- generated content (one or more registration audiovisual data samples Page 10) such as “registered AI- generated content” without relying on watermark detection (The audio deepfake detection engine may implement, for example, a neural network architecture or a GMM based architecture Page 9)
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Beauchesne in view of Sarkissian to incorporate the teachings of Tianxiang wherein the input content comprises deepfake content, and the system is configured to identify portions of the deepfake content that match registered AI- generated content without relying on watermark detection. By doing so determine the likelihood that audiovisual data contains deepfake content and the likelihood that the alleged identity of an individual within a video matches the identity of an expected or registered individual. Tianxiang Page 2
Claim 28 corresponds to claim 22 and is rejected accordingly
Regarding claim 25 Beauchesne in view of Sarkissian teaches the system of claim 1,
Beauchesne as modified does not fully disclose wherein comparing hash values comprises segmenting the input content into a plurality of parts, transforming each part into a part identifier using a one- way transformation function, and determining one or more portions of the input content that match a registered content group by calculating an edit distance between hash values of the input content portions and hash values of the registered content group.
Tianxiang teaches wherein comparing hash values (comparing the registered joint embedding with the inbound joint embedding. Page 18) comprises segmenting the input content (input audiovisual data Page 8) such as “input content” into a plurality of parts, (parsing and segmenting an audio signal or image data into frames and segments Page 8) transforming each part into a part identifier using a one- way transformation function, (and performing one or more transform functions, such as Short-time Fourier Transform (SFT) or Fast Fourier Transform (Page 8) and determining one or more portions (segments or frames of a given size Page 15)of the input content (input audiovisual data Page 8) such as “input content” that match a registered content group (an individual within a video matches the identity of an expected or registered individual Page 5, 7) (a talker or face in the inbound audiovisual data will match the voice or face of a registered talker. Page 15) by calculating an edit distance (the machine learning architecture extracts the inbound faceprint and the registered faceprint, and outputs a similarity score representing the distance between the inbound faceprint and the registered faceprint. Page 15 ) between hash values of the input content portions (joint embeddings for the audiovisual data 602 Page 16) and hash values of the registered content group (and the registered joint embeddings stored in the database. 626).)
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Beauchesne in view of Sarkissian to incorporate the teachings of Tianxiang segmenting the input content into a plurality of parts, transforming each part into a part identifier using a one- way transformation function, and determining one or more portions of the input content that match a registered content group by calculating an edit distance between of the input content portions and of the registered content group. By doing so the machine learning architecture 607 determines that the audiovisual data 602 is genuine or spoofed based on whether the final output score 626 satisfies a pre-configured threshold score. Tianxiang Page 16
Claim 31 corresponds to claim 25 and is rejected accordingly
Claims 27 and 33 are rejected under 35 U.S.C. 103 as being unpatentable over Beauchesne et al., (United States Patent Publication Number 2023/0143574) hereinafter Beauchesne in view of Sarkissian et al., (United States Patent Publication Number 2024/0311659) and in further view of Jana Dittman et al., (Using Cyrptographic and Watermarking Algorithms) hereinafter Dittman
Regarding claim 27 Beauchesne in view of Sarkissian teaches the system of claim 1,
Beauchesne as modified teaches the registry (fingerprints of known content 120 [0084]) such as “registry”
Beauchesne as modified does not fully disclose content generated by AI services that do not implement watermarking; and content generated by AI services that implement watermarking, wherein the system determines content origin independently of watermark detection to provide universal content verification.
Dittman teaches content generated by AI services that do not implement watermarking; (multimedia data derived from applying cryptographic mechanisms. Page 55) and content generated by AI services that implement watermarking, (Digital watermarking techniques based on steganographic systems can embed information directly into the media data Page 55) wherein the system determines content origin independently of watermark detection to provide universal content verification. (We can use public-key cryptosystems to generate and verify digital signatures.10,11 A digital sig
nature of an entity A (the signer) of data m depends on m and the private key of A. Each user can verify the authenticity of the signature created by A within a verification process using the public key of A Page 56)
It would have been prima facie obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified Beauchesne in view of Sarkissian to incorporate the teachings of Dittman wherein content generated by AI services that do not implement watermarking; and content generated by AI services that implement watermarking, wherein the system determines content origin independently of watermark detection to provide universal content verification. By doing so the integrity verification data is embedded in the media rather than appended to it. Possessing the appropriate secret key K, we can verify the watermark and evaluate whether the data was altered (particularly tampered) by checking the embedded information. Dittman Page 57
Claim 33 corresponds to claim 27 and is rejected accordingly
Conclusion
6. The prior art made of record and not relied upon is considered pertinent to
applicant's disclosure.
Kathrow et al., (United States Patent Number 6,263,348) teaches
"Fingerprint compare 242 compares the pair of fingerprints received from
fingerprint and content storage 230. If the fingerprints are identical, there is a high
probability that the Windows registry files corresponding to each of the pair of
fingerprints received by fingerprint compare 242 are identical. If either the hash
result or the characteristic of one fingerprint is different from that of the other
fingerprint, the files corresponding to these fingerprints are not identical Col 6 In 19
28
7. Any inquiry concerning this communication or earlier communications from the
examiner should be directed to Kweku Halm whose telephone number is (469) 295-
9144. The examiner can normally be reached on 7:30AM - 5:30PM Mon - Thur. If
attempts to reach the examiner by telephone are unsuccessful, the examiner's
supervisor, Sanjiv Shah can be reached on (571) 272-4098. The fax phone
number for the organization where this application or proceeding is assigned is 571-273-
8300.
Information regarding the status of an application may be obtained from the
Patent Application Information Retrieval (PAIR) system. Status information for published
applications may be obtained from either Private PAIR or Public PAIR. Status information
for unpublished applications is available through Private PAIR only. For more
information about the PAIR system, see http://pair-direct.uspto.gov. Should you have
questions on access to the Private PAIR system, contact the Electronic Business Center
(EBC) at 866-217-9197 (toll-free).
/KWEKU WILLIAM HALM/Examiner, Art Unit 2166
/SANJIV SHAH/Supervisory Patent Examiner, Art Unit 2166
Prosecution Insights