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 .
Response to Amendment
The Amendment filed 4/28/2026 has been entered. Claims 1, 8, and 15 are amended. Claims 1-20 are pending.
Specification
The disclosure is objected to because of the following informalities:
(1) in paragraphs [0004] and [0073], the phrase "comprising sets instructions executable" omits the word "of" and should read "comprising sets of instructions executable";
(2) in paragraph [0018], the phrase "Unless backdoor based watermarking, verification input queries do not contain ownership information" appears to intend "Unlike backdoor based watermarking";
(3) in paragraph [0063], the sentence "This component is further configured to a new watermarks to non-watermarked models" is grammatically incomplete and appears to intend "configured to add a new watermark to non-watermarked models"; and
(4) in paragraphs [0029], [0076], [0083], and [0090], the phrase "the determination of the subset of the plurality of groups fairness bias" contains the superfluous words "fairness bias" and should read "the determination of the subset of the plurality of groups," consistent with the first sentence of paragraph [0029].
Appropriate correction is required.
Claim Objections
Claims 1-2, 4-6, 8-9, 11-13, 15-16, and 18-20 are objected to because of the following informalities:
Claim 1 recites "storing computer program code comprising sets instructions executable by the one or more processors," which appears to omit a word and should read "comprising sets of instructions," consistent with claim 8.
Claim 1 recites "one or more machine-readable medium coupled to the one or more processors," which is inconsistent in number and should read "one or more machine-readable media."
Claims 1, 8, and 15 each recite "input data belonging to subset of the plurality of groups," which appears to omit the article "the" and should read "input data belonging to the subset of the plurality of groups," consistent with the immediately following recitation "input data belonging to the subset."
Claims 5, 12, and 19 each recite "the clustering algorithm"; however, claims 1, 8, and 15, from which claims 5, 12, and 19 respectively depend, recite "a secret clustering algorithm." For consistent terminology and proper antecedent basis, "the clustering algorithm" should read "the secret clustering algorithm."
Claims 2, 9, and 16 each recite "a portion of the plurality results that correspond to the subset," which appears to omit a word and should read "a portion of the plurality of results," consistent with the earlier recitation in each claim of "a plurality of results."
Claims 4, 11, and 18 each recite "the determination of the subset of the plurality of groups fairness bias is based on ordering the plurality of groups based on a corresponding fairness bias of each group." The term "fairness bias" immediately following "the plurality of groups" appears to be superfluous and renders the recitation grammatically incomplete; for clarity the limitation should read "the determination of the subset of the plurality of groups is based on ordering the plurality of groups based on a corresponding fairness bias of each group." This objection is directed only to the informality; the metes and bounds of the limitation are otherwise ascertainable when the claim is read in light of the specification, for example at paragraphs [0029] and [0048].
Claims 6, 13, and 20 each recite "the modifying labels for data in the subset is based on a sensitivity bias," which appears to omit a word and should read "the modifying of labels for data in the subset," consistent with the modifying step recited in claims 1, 8, and 15, respectively.
Appropriate correction is required.
Claim Rejections - 35 USC 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
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.
Claims 1-3, 5-10, 12-17, and 19-20 are rejected under 35 U.S.C. 103 as being unpatentable over Jagielski et al. (hereinafter Jagielski) "Subpopulation Data Poisoning Attacks" (2021) in view of Solans et al. (hereinafter Solans) "Poisoning Attacks on Algorithmic Fairness" (2020) and further in view of Adi et al. (hereinafter Adi) "Turning Your Weakness Into a Strength: Watermarking Deep Neural Networks by Backdooring" (2018).
Regarding independent claim 1, Jagielski teaches a computer system, comprising: one or more processors; one or more machine-readable media coupled to the one or more processors and storing computer program code comprising sets of instructions executable by the one or more processors to (Jagielski: page 3105, Section 2.1 Neural Networks Background, "The goal of a learning algorithm A, when given a dataset D is to return a model f with parameters"; Jagielski sets forth a computer-implemented machine learning framework in which a learning algorithm (computer program code) is executed to train and evaluate neural network models, the processors (one or more processors) and machine-readable storage (one or more machine-readable media) being the conventional computing components that carry out and store the disclosed training): obtain an original set of labeled data including original data and an original set of labels classifying each piece of the original data (Jagielski: page 3105, Section 2.1, "Consider a training set of n examples" in which there is "each feature vector" and a corresponding "label"; the training set (an original set of labeled data) is composed of feature vectors (original data) each paired with a label (an original set of labels) that classifies the corresponding data point);
cluster the original set of labeled data into a plurality of groups using a clustering algorithm (Jagielski: page 3108, Section 4.1.2 ClusterMatch, "replaces the need for annotation with clustering to identify subpopulations of interest" and "By identifying natural clusters in the data, one can compromise the model for one cluster but not elsewhere" and "For clustering, we use KMeans, but any procedure for generating meaningful clusters on a given dataset should work"; Jagielski applies a clustering algorithm (a clustering algorithm), namely KMeans clustering of the training data, that partitions the data into a plurality of natural clusters (a plurality of groups));
determine a subset of the plurality of groups (Jagielski: page 3108, Section 4.1 Subpopulation Selection, "The adversary first needs to identify their target subpopulation"; page 3108, Section 4.1.2 ClusterMatch, "an adversary can generate a clustering and identify a cluster that is both aligned with their goals and will be easy to attack"; the selected target subpopulation, i.e., an identified one of the clusters, is the target subpopulation (a subset of the plurality of groups));
modify labels for data in the subset of the plurality of groups to obtain modified labels for the data in the subset (Jagielski: page 3109, Section 4.2.1 Label Flipping, "adding these to the training set with a label t different from the original one c"; the label-flipping generation replaces, for points drawn from the selected subpopulation, the original label with a different label (a flipped label), thereby producing modified labels for the data in the subset);
train a machine learning model based on the subset of data labeled using the modified labels and the original set of data outside of the subset labeled using the original set of labels, wherein the machine learning model exhibits when classifying input data belonging to the subset of the plurality of groups (Jagielski: pages 3104-3105, Section 1.1 Our Contributions, "compromise the performance of a classifier on a particular subpopulation of interest, while maintaining unaltered its performance for the rest of the data"; the classifier (a machine learning model) is trained on the flipped-label subpopulation data together with the unaltered remainder of the training data, so that at inference the trained classifier exhibits the inserted differential behavior on inputs drawn from the target subpopulation (input data belonging to the subset) while behaving normally elsewhere);
Jagielski does not expressly teach the modifying of the labels for data in the subset inserting fairness bias into the subset; and machine learning model exhibits the fairness bias when classifying.
However, Solans teaches the modifying of the labels for data in the subset inserting fairness bias into the subset (Solans: page 1, Abstract, "poisoning attacks against algorithmic fairness" and "introducing classification disparities among different groups in the data"; Solans: page 2, Introduction, "an attacker may be able to introduce algorithmic discrimination by developing a novel poisoning attack" whose purpose is to "create or increase a disadvantage against a specific group of individuals or samples"; Solans modifies training data so that the trained model exhibits a targeted algorithmic discrimination (a fairness bias) against a selected group, which is the fairness bias inserted into the subset); and machine learning model exhibits the fairness bias when classifying (Solans: page 7, Section 2.2 “Finally, in Fig. 2, we show how our poisoning attach modifies the decision function of a linear classifier to worsen algorithmic fairness”; Solans linear classifier exhibits the targeted algorithmic discrimination when classifying).
Because Jagielski and Solans are analogous art and within the same field of endeavor, specifically adversarial machine learning through training-data manipulation of classifiers, and are reasonably pertinent to the same problem of controlling a trained classifier's behavior on a selected portion of the input space by modifying training labels, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to configure the label modification of Jagielski's selected subpopulation so as to insert a fairness bias, as taught by Solans, with a reasonable expectation of success, such that the modifying of the labels for the data in the subset inserts a fairness bias into the subset, to teach the modifying of the labels for data in the subset inserting fairness bias into the subset; wherein the machine learning model exhibits the fairness bias when classifying input data belonging to the subset of the plurality of groups. This modification would have been motivated by the desire to produce a controlled, group-specific discrimination in the classifier's outputs (Solans: page 2, Introduction).
Jagielski and Solans do not expressly teach clustering using a secret clustering algorithm, or the exhibiting of the fairness bias for input data belonging to the subset being a watermark of the machine learning model that was trained using the modified labels for the subset.
However, Adi teaches clustering using a secret clustering algorithm and the exhibiting of the fairness bias for input data belonging to the subset being a watermark of the machine learning model that was trained using the modified labels for the subset (Adi: page 5, Section 3 Defining Watermarking, "a first algorithm to generate the secret marking key mk which is embedded as the watermark"; page 2, Introduction, "the ability of an operator to train a model to deliberately output specific (incorrect) labels for a particular set of inputs T"; page 1, Abstract, "we present an approach for watermarking Deep Neural Networks in a black-box way"; page 2, Introduction, "we turn this curse into a blessing by reducing the task of watermarking a Deep Neural Network to that of designing a backdoor for it"; page 9, Section 5.1, "To embed the watermark, we optimize the models using both training set and trigger set"; Adi generates a secret marking key, thus supplying the claimed secret character of the clustering algorithm (a secret clustering algorithm), that is an owner-held parameter known only to the model owner and that governs which inputs are deliberately marked; Adi deliberately trains the model on that owner-held subset so that the resulting exhibited behavior, trained to deliberately output specific (incorrect) labels for a particular set of inputs T (the exhibiting of the fairness bias for input data belonging to the subset), is a watermark of the machine learning model (being a watermark of the machine learning model that was trained using the modified labels for the subset) usable to verify ownership).
Because Jagielski, in view of Solans, and Adi are analogous art and within the same field of endeavor, specifically the training and protection of machine learning models, and are reasonably pertinent to the same problem of embedding an owner-controlled, verifiable behavior into a trained model, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to keep the clustering algorithm of Jagielski secret and known only to the model owner and to treat the resulting inserted fairness bias as a watermark, as taught by Adi, with a reasonable expectation of success, such that the data is clustered using a secret clustering algorithm and the exhibited fairness bias serves as a watermark of the model, to teach cluster the original set of labeled data into a plurality of groups using a secret clustering algorithm and the exhibiting of the fairness bias for input data belonging to the subset being a watermark of the machine learning model that was trained using the modified labels for the subset. This modification would have been motivated by the desire to provide a tracking mechanism for model ownership (Adi: page 1, Introduction). Also, Adi teaches that deliberately training a model to produce assigned outputs on a secret owner-held subset yields an ownership watermark that combines readily with existing training algorithms, giving the skilled artisan a reasonable expectation that the same secret-subset training used by Jagielski and Solans would function as such a watermark. See KSR Int'l Co. v. Teleflex Inc., 550 U.S. 398 (2007); MPEP 2143.01.
Regarding dependent claim 2, Jagielski, in view of Solans and Adi, teach the computer system of claim 1, wherein the computer program code further comprises sets of instructions executable by the one or more processors to: send a plurality of inference queries to a machine learning service to obtain a plurality of results, the plurality of inference queries including query data belonging to the subset of the plurality of groups (Adi: page 6, Section 4, Verify algorithm, "test that Classify(t(i),M)" for the trigger-set inputs t(i); Adi verifies a suspect model by submitting the owner-held subset inputs (query data belonging to the subset) to the model offered by a service provider (a machine learning service) and reading the returned classifications (a plurality of results)); determine whether a portion of the plurality of results that correspond to the subset exhibit the fairness bias (Adi: page 5, Section 3, "a third algorithm to verify if a watermark is present in a model or not"; Adi checks whether the results returned for the subset inputs match the deliberately assigned behavior, i.e., exhibit the inserted bias); identify the watermark of the machine learning model based on the determination of the fairness bias (Adi: page 1, Abstract, "we present an approach for watermarking Deep Neural Networks in a black-box way"; when the results for the subset inputs exhibit the assigned behavior, the watermark is identified in a black-box manner from that determination); and determine that the machine learning service uses the machine learning model based on the watermark (Adi: page 1, Abstract, "a tracking mechanism to identify models as the intellectual property of a particular vendor is necessary"; identification of the watermark establishes that the service is using the owner's model).
Regarding dependent claim 3, Jagielski, in view of Solans and Adi, teach the computer system of claim 1, wherein the fairness bias is based on a disparate impact metric (Solans: page 4, Section 2.1, "we consider the disparate impact criterion" and "Disparate impact thus measures the ratio between the fractions of unprivileged and privileged samples that are assigned to the positive class"; the fairness bias introduced by Solans is defined and measured by the disparate impact criterion, which is a disparate impact metric).
Regarding dependent claim 5, Jagielski, in view of Solans and Adi, teach the computer system of claim 1, wherein the secret clustering algorithm is unique and deterministic (Jagielski: page 9, Section 4.1.2 ClusterMatch, "For clustering, we use KMeans, but any procedure for generating meaningful clusters on a given dataset should work"; the KMeans clustering procedure (a clustering algorithm) yields a fixed and reproducible partition of the data for a given dataset and initialization, and is therefore unique and deterministic, its secrecy being provided by Adi as set forth for claim 1).
Regarding dependent claim 6, Jagielski, in view of Solans and Adi, teach the computer system of claim 1, wherein the modifying of labels for data in the subset is based on a sensitivity bias (Jagielski: page 3109, Section 4.2.1 Label Flipping, "the adversary uses a poisoning rate" relative to the subpopulation to add a corresponding number of poisoned points; the amount of label modification is governed by a poisoning rate, that is, a sensitivity parameter that sets the proportion of modified labels in the subpopulation, which is the sensitivity bias on which the modifying of labels is based).
Regarding dependent claim 7, Jagielski, in view of Solans and Adi, teach the computer system of claim 1, wherein the machine learning model is a binary classifier or a multi-class classifier (Jagielski: page 3105, Section 2.1, "we consider multiclass classification" tasks in which "a K-class problem" is addressed; the classifier (a machine learning model) is a multi-class classifier, and only one member of the recited alternative group need be taught).
Regarding claims 8-10 and 12-14, these are non-transitory computer-readable medium claims that are substantially the same as the computer system of claims 1-3 and 5-7, respectively. Thus, claims 8-10 and 12-14 are rejected for the same reasons as claim 1-3 and 5-7. In addition, Jagielski teaches a non-transitory computer-readable medium storing computer program code comprising sets of instructions to carry out the recited operations (Jagielski: page 3, Section 2.1, "The goal of a learning algorithm A, when given a dataset D is to return a model f with parameters"; the disclosed learning algorithm is embodied as computer program code (computer program code comprising sets of instructions) stored on machine-readable storage (a non-transitory computer-readable medium) and executed to perform the training and evaluation set forth above).
Regarding claims 15-17 and 19-20, these are computer-implemented method claims that are substantially the same as the computer system of claims 1-3 and 5-6, respectively. Thus, claims 15-17 and 19-20 are rejected for the same reasons as claims 1-3 and 5-6. In addition, Jagielski teaches a computer-implemented method, comprising the recited steps (Jagielski: page 8, Section 4.1, "The adversary first needs to identify their target subpopulation"; the disclosed subpopulation poisoning is a computer-implemented method performed by executing the learning algorithm on a computing system).
Claims 4, 11, and 18 are rejected under 35 U.S.C. 103 as being unpatentable over Jagielski, in view of Solans and Adi, as applied in the rejections of claims 1, 8, and 15 above, and further in view of Suya et al. (hereinafter Suya) "Model-Targeted Poisoning Attacks with Provable Convergence" (2021).
Regarding dependent claim 4, Jagielski, in view of Solans and Adi, teach all the elements of claim 1.
Jagielski, Solans, and Adi do not expressly teach wherein the determination of the subset of the plurality of groups is based on ordering the plurality of groups based on a corresponding fairness bias of each group.
However, Suya teaches determining a subset of clustered groups based on ordering the plurality of groups based on a corresponding metric of each group (Suya: page 6, Section 5 Datasets and Subpopulations, "We identify the subpopulations for the Adult dataset using k-means clustering techniques (ClusterMatch in Jagielski et al. (2019)) to obtain different clusters"; Suya clusters the training data into twenty subpopulations (the plurality of groups) by k-means clustering; page 6, Section 5 Datasets and Subpopulations, "From the 20 subpopulations obtained, we select three subpopulations with the highest test accuracy on the clean model"; the three selected subpopulations (a subset of the plurality of groups) are determined by ranking the twenty subpopulations on the test accuracy of each subpopulation on the clean model, and selecting the subpopulations with the highest value of a per-subpopulation metric is determining the subset based on ordering the groups based on the corresponding metric of each group).
Because Jagielski, in view of Solans and Adi, and Suya are analogous art, each being within the same field of endeavor of machine learning data poisoning and the selection of subpopulations of training data, and each being reasonably pertinent to the problem of choosing which clustered group to target so that a trained model exhibits a controlled behavior, accordingly, it would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to apply the metric-ranked subpopulation selection of Suya to the combination of Jagielski, Solans, and Adi, using the per-group classification-disparity score of Solans as the ranking metric, with a reasonable expectation of success, such that the target subpopulation is determined by ordering the clusters according to the fairness-bias score of each cluster and selecting the highest-ranked cluster, thereby teaching wherein the determination of the subset of the plurality of groups is based on ordering the plurality of groups based on a corresponding fairness bias of each group. This modification would have been motivated by Suya's own observation that subpopulations may be chosen by the very group characteristics on which a fairness bias is measured (Suya: page 6, Section 5 Datasets and Subpopulations) and by the desire to single out the group against which the inserted disadvantage is most effective (Solans: page 3, Introduction).
Regarding dependent claim 11, it is a non-transitory computer-readable medium claim that is substantially the same as the computer system of claim 4. Thus, claim 11 is rejected for the same reason as claim 4.
Regarding dependent claim 18, it is a computer-implemented method claim that is substantially the same as the computer system of claim 4. Thus, claim 18 is rejected for the same reason as claim 4.
Response to Arguments
Applicant's claim amendments and Remarks filed 04/28/2026 traversing the 35 U.S.C. 103 rejections have been fully considered but they are not persuasive.
Applicant first argues (Remarks, pages 7-8) that neither Adi, Jagielski, Solans, nor any other cited art, alone or in combination, teaches, suggests, or otherwise renders obvious the amended independent claims, reciting the limitations of claim 15 and emphasizing the newly added "secret clustering algorithm."
Examiner respectfully disagrees. This argument is not persuasive for three reasons. First, apart from the secret clustering algorithm limitation, the argument recites the claim language and asserts in general terms that the art does not render it obvious, without specifically pointing out how the language of the claims patentably distinguishes them from the references; a general allegation that the claims define a patentable invention is not a proper traverse of the detailed limitation-by-limitation mapping set forth above. See 37 CFR 1.111(b); MPEP 707.07(f). Second, as to the secret clustering algorithm, the rejection does not assert that any single reference expressly teaches a secret clustering algorithm; the rejection states that gap expressly and fills it through the combination. Adi expressly teaches generating and holding a secret parameter that governs which inputs carry the embedded behavior (Adi: page 5, Section 3 Defining Watermarking, "a first algorithm to generate the secret marking key mk which is embedded as the watermark"), and the rejection articulates a reference-grounded reason to keep the clustering algorithm of Jagielski secret and known only to the model owner, namely Adi's ownership-tracking purpose. Nonobviousness cannot be established by attacking the references individually where the rejection is based on a combination of references. See In re Keller, 642 F.2d 413 (CCPA 1981); In re Merck & Co., 800 F.2d 1091 (Fed. Cir. 1986); MPEP 2145(IV). Third, the broadest reasonable interpretation of "secret" is supplied by Applicant's own specification, which explains that the clustering algorithm "may be secret and only known to the model owner" (specification paragraphs [0020], [0024], [0028], [0043]); under that interpretation, the combination in which the model owner performs the clustering of Jagielski privately under the secret-parameter discipline of Adi is a clustering algorithm known only to the model owner, which is all the claim requires.
Applicant next argues (Remarks, page 8) that Adi "merely discloses" a cryptographic modeling of the tasks of watermarking and backdooring constructed in a black-box way.
Examiner respectfully disagrees. This argument is not persuasive. Adi is not relied upon for the clustering, label-modification, or training operations, which are supplied by Jagielski and Solans as set forth above; Adi is relied upon for the secret owner-held parameter, for treating the deliberately trained behavior as a watermark, and for black-box verification. The sentence Applicant quotes does not negate Adi's further teachings, including "we turn this curse into a blessing by reducing the task of watermarking a Deep Neural Network to that of designing a backdoor for it" (page 2, Introduction) and "To embed the watermark, we optimize the models using both training set and trigger set" (page 9, Section 5.1). A reference may be relied upon for all that it teaches, and the cited disclosure is not limited to the passage Applicant quotes. See MPEP 2123; In re Keller, 642 F.2d 413 (CCPA 1981).
Applicant similarly argues (Remarks, page 8) that Jagielski "merely discloses" ClusterMatch, which replaces the need for annotation with clustering to identify subpopulations of interest.
Examiner respectfully disagrees. This argument is not persuasive because the passage Applicant quotes is the very disclosure applied in the rejection: ClusterMatch clusters the training data into a plurality of natural clusters (the plurality of groups), identifies a target cluster (the subset), and Jagielski flips labels for points of that cluster and trains the model so that it is compromised on the cluster while performing normally elsewhere. Jagielski is not relied upon alone for the fairness character of the inserted behavior, which Solans supplies, or for the secrecy and watermark character, which Adi supplies. See MPEP 2145(IV).
Applicant further argues (Remarks, page 8) that Solans "merely discloses" an optimization framework for poisoning attacks against algorithmic fairness.
Examiner respectfully disagrees. This argument is not persuasive because the quoted disclosure is the teaching applied in the rejection: a poisoning attack "aimed at introducing classification disparities among different groups in the data" is the insertion of a fairness bias by modification of the training data, and Solans further explains that the attacker's purpose is to "create or increase a disadvantage against a specific group of individuals or samples" (page 2, Introduction). Solans is not relied upon for clustering, secrecy, or watermark verification. See MPEP 2145(IV).
Applicant finally argues (Remarks, pages 8-9) that dependent claims 2-7, 9-14, and 16-20 are allowable at least by virtue of their dependency from claims 1, 8, and 15.
Examiner respectfully disagrees. Because the rejections of independent claims 1, 8, and 15 are maintained as restructured, the dependent claims fall with the independent claims. Claims 4, 11, and 18 are additionally rejected in view of Suya as set forth above. No separate argument directed to the added limitation of any dependent claim has been presented.
For the reasons set forth above, Applicant's arguments have been fully considered but are not persuasive. The rejections under 35 U.S.C. 103 are maintained as restructured in this Office Action.
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to KUANG FU CHEN whose telephone number is (571)272-1393. The examiner can normally be reached M-F 9:00-5:30pm ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Jennifer Welch can be reached on (571) 272-7212. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/KC CHEN/Primary Patent Examiner, Art Unit 2143
Prosecution Insights