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
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and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114.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 01/06/2026 has been entered.
Response to Arguments
Applicant’s arguments with respect to claim(s) 1 and 11 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument.
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.
Claim(s) 1, 3-4, 6-9, 11, 13-14, and 16-19 are rejected under 35 U.S.C. 103 as being unpatentable over Subramanya et al. (US-20240195701-A1) in view of Donovan (US-20210299576-A1), Dzierzanowski et al. (US-20220237565-A1), and Bhargava et al. (US-20200005168-A1).
Regarding Claim 1,
Subramanya (US 20240195701 A1) teaches a method, comprising:
for each artificial intelligence (AI) ethics pillar in a group of Al ethics pillars, storing, in a datastore, context data, and the context data is determined using context rules concerning that Al ethics pillar (para [0098]-[0099] “The AI Trust Engine translates the AI QoT Intent/Class Identifier into AI Trustworthy (i.e., Fairness, Robustness and Explainability) requirements and sends it to the AI Trust Manager of the AI Pipeline over T2 interface. The AI Trust Manager configures, monitors and measures AI Trustworthy requirements (i.e., trust mechanisms and trust metrics) for at least one of AI Data Source Manager, AI Training Manager and AI Inference Manager over T3, T4 and T5 interfaces respectively.” AI Trustworthy requirements (i.e., context rules) for fairness, robustness, explainability (i.e., AI ethics pillars) are used to determine measured metrics (i.e., context data). para [0158] “The means for obtaining 235 obtains an actual value of the at least one of the fairness, the explainability, and the robustness from the data source manager, and/or the training manager, and/or the inference manager, respectively (S235).” Values for fairness, explainability, and robustness (i.e., context data). para [0107] “Trust Knowledge Database of AI Trust Engine stores all trust related metrics or explanations or artifacts received from AI Trust Managers of various AI Pipelines.” Context data is stored.);
storing, in the datastore, the context rules as minimum context requirements (para [0098]-[0099] And para [0100] “The AI Trust Manager may compare the received actual values with the corresponding requirement. If an actual value does not fulfil the corresponding requirement, the AI Trust Manager may issue an error notice.” The trust level requirements are minimum requirements. para [0152] “For example, the apparatus knows from a database entry and/or from a received message comprising the trust level requirement (or being related to the message comprising the trust level requirement) that the artificial intelligence pipeline is involved in providing the service, and/or that the trust manager is related to the artificial intelligence pipeline.” Trust level requirements are stored in a database.);
ensuring that an assessment mechanism is able to access, and assess, the context data for each Al ethics pillar (para [0100] “The AI Trust Manager may compare the received actual values with the corresponding requirement. If an actual value does not fulfil the corresponding requirement, the AI Trust Manager may issue an error notice.” The AI trust manager of the AI pipeline may assess the measured values (i.e., context data) by determining whether it meets the requirement. para [0115] “The AI Trust Manager maps the Fairness, Explainability and Robustness AI Trustworthy metrics/explanations received from the AI Trust Engine to AI Trustworthy mechanisms (e.g., post-processing fairness, post-modelling explainability) required to meet those requirements at Inference stage of the AI/ML Pipeline and configures them for AI Inference Manager through T5 interface.” AI pipeline (i.e., assessment mechanism).).
Subramanya does not explicitly disclose
receiving, by the datastore, a request from a user to register an asset in the datastore;
when user-supplied context information for the asset meets ethical requirements specified by the context rules, registering the asset in the datastore; and
wherein the ethical requirements, which are specific to the registered asset, are versioned and tracked;
storing, in the datastore, results of an assessment process performed concerning the asset with respect to the Al ethics pillars, wherein the assessment process comprises at least one of: (i) a self-measurement process performed by the user or (ii) a peer assessment process.
However, Donovan (US 20210299576 A1) teaches
receiving, by the datastore, a request from a user to register an asset in the datastore (para [0037] “In the example shown in FIG. 3A, the gaming application takes the form of a social media question and answer game where players/users can post questions including text and/or other media for review.” A user can request to post (i.e., register) social media content (i.e., asset) for review.);
when user-supplied context information for the asset (para [0024] And para [0059] Game data (i.e., user supplied context information.) meets ethical requirements specified by the context rules, registering the asset in the datastore (para [0026] “Once the moral insights platform 130 generates/trains the AI moral insight model, the moral insights platform 130 receives new media content 122 from a media source 140 for evaluation by the model. The moral insights platform 130 utilizes the AI moral insight model to generate predicted moral score data 124 associated with the new media content 122 corresponding to the at least one moral index. This predicted moral score data 124 is sent to the media source 140 and can be compared with one or more moral index thresholds, for example, to block content with that compares unfavorably to such threshold(s) and/or to validate, certify or allow content that compares favorably to such threshold(s).” New media content (i.e., asset). Validate, Certify, or allow (i.e., register).); and
Subramanya and Donovan are analogous because they are directed to the same field of endeavor of ethical artificial intelligence.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the ethical AI model of Subramanya with the moral index thresholds of Donovan.
Doing so would allow for blocking or allowing content based on ethical and moral considerations (Donovan para [0049]).
Dzierzanowski (US 20220237565 A1) teaches
wherein the ethical requirements, which are specific to the registered asset, are versioned and tracked (para [0053] “In various embodiments, versions, and version control of frameworks, policies, legal and regulatory requirements, and/or associated materials are systematically captured and maintained.” Para [0085]-[0086] “One such element may be explanation capabilities. These elements and others may be combined with domain specific assessments of AI/ML projects developed via system 100. Knowledge catalog 200 includes one or more subsystem processes such as, for example, project attributes and various embodiments for structuring entity policy, ethical guidelines, and governance processes 202, Algorithmic Impact Assessments (MA) 204, Social Impact Assessments (SIA) 206, Privacy Impact Assessments (PIA) 208, and Model Risk Management (MRM) 210 processes for capturing requirements.”. Para [0097]-[0098] This section shows a AI-regulatory framework with requirements and the date, time, author status, etc. for an AI asset. Para [0121] “In various embodiments, the hierarchy 450 serves as a basis for the calculation includes AI/ML model source code 452, entity policy statements and or other governance documents 454 associated with the entity's project, AI/ML model validation and verification parameters 456 created during development and testing, relevant forms of project documentation 458, and data 460 which may include data utilized in project development for training 462, quality assurance 464 and/or production snapshots data 466.” Para [0142]-[0143]).
Subramanya and Dzierzanowski are analogous because they are directed to the same field of endeavor of ethical AI.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the ethical AI model of Subramanya with the knowledge catalog of Dzierzanowski.
Doing so would allow for effectively capturing multiple domain and sector-specific assessments, design decisions and establishing evidence grade accountability with immutability is desirable for monitoring, model management, model-based longitudinal studies, algorithmic audit, regulatory, compliance, risk management, reputational risk, financial, ethical, equity, discovery, legal actions, and societal impacts pertaining to AI/ML models (Dzierzanowski para [0003]).
Bhargava (US 20200005168 A1) teaches
storing, in the datastore (para [0031] “In a further embodiment of the system, the MTI database stores a plurality of MTIs generated at a plurality of different times for a plurality of different AI processes.”), results of an assessment process performed concerning the asset with respect to the Al ethics pillars (para [0038] “It makes major, novel contributions, to the emerging concepts and practice of “Ethical AI” and “Principled AI”. In particular, it may provide a flexible and transparent methodology and platform that allows for a quantified measurement of “Machine Trust”, referred to as a machine trust index (MTI) against a transparent, configurable and auditable set of measures and criteria. The MTI allows for relative comparisons of different algorithms and models once a common measurement methodology is determined.”), wherein the assessment process comprises at least one of: (i) a self-measurement process performed by the user (para [0058] “The audit module 306 may provide functionality for template- and script-driven auditing, in which the auditor (e.g., Lead) is guided through the steps of the MTI scoring, the specific sequence and questions conditioned both on the auditor's settings and on the data shared from the developer module of that project.”) or (ii) a peer assessment process (para [0052] “audit information and functions, for some (again typically small) number of leads and other auditing, reviewing, scoring personnel who rate the developed models and code according to the MTI or review the models, code and score on behalf of users, regulatory bodies or others;”).
Subramanya and Bhargava are analogous because they are directed towards evaluating ethical trustworthiness of AI models.
It would have been obvious to one or ordinary skill in the art before the effective filing date to modify the method of evaluating the ethical trustworthiness of Subramanya with the MTI scoring of Bhargava.
Doing so would allow for collecting machine trust data breadth, depth and over time to allow for deep learning techniques to evolve and improve new trust tools used in analyzing AI processes and insights provided (Bhargava para [0038]).
Regarding Claim 3,
Subramanya, Donovan, Dzierzanowski, and Bhargava teach the method as recited in claim 1. Subramanya further teaches wherein the assessment mechanism is a machine-based assessment mechanism (para [0098] “The AI Trust Engine translates the AI QoT Intent/Class Identifier into AI Trustworthy (i.e., Fairness, Robustness and Explainability) requirements and sends it to the AI Trust Manager of the AI Pipeline over T2 interface. The AI Trust Manager configures, monitors and measures AI Trustworthy requirements (i.e., trust mechanisms and trust metrics) for at least one of AI Data Source Manager, AI Training Manager and AI Inference Manager over T3, T4 and T5 interfaces respectively.”)
Regarding Claim 4,
Subramanya, Donovan, Dzierzanowski, and Bhargava teach the method as recited in claim 1. Subramanya further teaches wherein the asset is an algorithm (para [0106] “For example, for an autonomous driving service, since the risk level is very high, the trust level requirements (e.g., fairness, explainability, robustness) for an AI/ML model (e.g., proactive network-assisted lane changing) is very high. On the other hand, for a movie streaming service, since the risk level is mostly low, the trust requirements for an AI/ML model (e.g., proactive caching of video chunks from cloud to edge) is also very low.” AI/ML model (i.e., algorithm).).
Regarding Claim 6,
Subramanya, Donovan, Dzierzanowski, and Bhargava teach the method as recited in claim 1.
Subramanya further teaches an assessment process performed by a machine learning algorithm (para [0098] “The AI Trust Engine translates the AI QoT Intent/Class Identifier into AI Trustworthy (i.e., Fairness, Robustness and Explainability) requirements and sends it to the AI Trust Manager of the AI Pipeline over T2 interface. The AI Trust Manager configures, monitors and measures AI Trustworthy requirements (i.e., trust mechanisms and trust metrics) for at least one of AI Data Source Manager, AI Training Manager and AI Inference Manager over T3, T4 and T5 interfaces respectively.”)
Bhargava further teaches
wherein the assessment process comprises all of: the self-measurement process performed by the user (para [0058] “The audit module 306 may provide functionality for template- and script-driven auditing, in which the auditor (e.g., Lead) is guided through the steps of the MTI scoring, the specific sequence and questions conditioned both on the auditor's settings and on the data shared from the developer module of that project.”); and the peer assessment process (para [0052] “audit information and functions, for some (again typically small) number of leads and other auditing, reviewing, scoring personnel who rate the developed models and code according to the MTI or review the models, code and score on behalf of users, regulatory bodies or others;”).
It would have been obvious to one or ordinary skill in the art before the effective filing date to modify the method of evaluating the ethical trustworthiness of Subramanya with the MTI scoring of Bhargava.
Doing so would allow for collecting machine trust data breadth, depth and over time to allow for deep learning techniques to evolve and improve new trust tools used in analyzing AI processes and insights provided (Bhargava para [0038]).
Regarding Claim 7,
Subramanya, Donovan, Dzierzanowski, and Bhargava teach the method as recited in claim 6. Donovan further teaches further comprising generating, for one or more of the Al ethics pillars, a respective tally score based on the context data for the Al ethics pillar and based on results of the assessment processes (para [0024] “Furthermore, the game data 118 sent from the gaming devices to the game platform 125 can include other game output, including but not limited to the reviewed media content as well as moral score data associated with moral index data that indicates one or more moral indices or factors associated with each instance of the reviewed media content. The game data 118 can be sorted, formatted and/or otherwise processed by the game platform 125 to generate the training data 120 that includes the reviewed media content and the corresponding moral score data.” Gaming data (i.e., context data) may be assess by ML models to generate scores (i.e., tally scores) para [0038] “The moral indices presented include factors for immoral/moral, betrayal/loyalty, degradation/sanctity, harm/care, cheating/fairness and subversion/authority, however, in other examples, some subset of one or more of these factors could likewise be employed as well as the use of other moral indices not expressly shown.” Moral indices (i.e., AI ethics pillars).).
Subramanya and Donovan are analogous because they are directed to the same field of endeavor of ethical artificial intelligence.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the ethical AI model of Subramanya with the moral scores of Donovan.
Doing so would allow for blocking or allowing content based on ethical and moral considerations (Donovan para [0049]).
Regarding Claim 8,
Subramanya, Donovan, Dzierzanowski, and Bhargava teach the method as recited in claim 1. Subramanya further teaches further comprising tracing, over time, the asset (para [0094] “In contrast, some example embodiments of the invention provide a framework to configure, measure and/or monitor the trustworthiness (i.e., fairness, robustness and explainability) of the designed/deployed AI/ML model over its entire lifecycle.” AI/ML model is monitored (i.e., traced).), context data (para [0098] “The AI Trust Manager configures, monitors and measures AI Trustworthy requirements (i.e., trust mechanisms and trust metrics) for at least one of AI Data Source Manager, AI Training Manager and AI Inference Manager over T3, T4 and T5 interfaces respectively.”), and results of assessments of the asset (para [0099] metrics/explanations/artifacts (i.e., assessment results) are measured or collected periodically. para [0100] The AI Trust Manager may compare the received actual values with the corresponding requirement. If an actual value does not fulfil the corresponding requirement, the AI Trust Manager may issue an error notice. For example, the error notice may be stored in a logfile, and/or the AI Trust Manager may provide the error notice to the AI Trust Engine. Error notice is an example of an assessment result that is stored.).
Donovan further teaches
versioning and tracing, over time, the asset (para [0034] “The game data 118 can include, for example, a current version of a gaming application 248 that is presented to the gaming devices for play.” And para [0068] “The ethical AI development platform 800 also provides access to a version control repository 812, such as a Git repository or other version control system for storing and managing a plurality of versions of the training dataset and the AI model.”),
Subramanya, Donovan, Dzierzanowski, and Bhargava are analogous because they are directed to the same field of endeavor of ethical artificial intelligence.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the ethical AI model of Subramanya with the versioning of Donovan.
Doing so would allow for controlling, storing, and managing a plurality of versions of the asset (Donovan para [0049]).
Regarding Claim 9,
Subramanya, Donovan, Dzierzanowski, and Bhargava teach the method as recited in claim 1. Subramanya further teaches further comprising:
receiving, from a requestor, a request to use the asset (para [0118]-[0119] A user requests a service associated with a AI/ML model (i.e., asset).);
When the context of the asset complies with the context data of an organization to which the requestor belongs (para [0118] “Action 1: Customer requests for a service via Intent request (i.e., the request for the service comprises a (business) intent).” Business (i.e., organization).), making the asset available to the requestor (para [0124] “Actions 8-9: AI Training Manager fetches the RRC measurement data from AI Data Source Manager and then trains the AI model with requested levels of Trust and Performance. For example, in case a neural network is being used for proactive mobility management, a surrogate explainable model might also be trained to satisfy the required explainability levels. It then pushes the model Trust metrics/explanations/artifacts to AI Trust Manager.” The AI/ML model (i.e., asse). must meet requirements before being deployed para [0159] “The means for comparing 240 compares at least one of the received actual values with the corresponding requirement (S240). If the at least one of the actual values does not fulfill the corresponding requirement (S240=no), the means for issuing 250 issues an error notice (S250), such as an alarm.” If measured values (i.e., context data) does not meet the requirements, an error notice may be issued and corrective action may be taken to ensure the AI/ML model meets the requirements.).
Regarding Claim 11,
Claim 11 is the non-transitory storage medium corresponding to the method of claim 1. Claim 11 is substantially similar to claim 1 and is rejected on the same grounds.
Regarding Claim 13,
Claim 13 is the non-transitory storage medium corresponding to the method of claim 3. Claim 13 is substantially similar to claim 3 and is rejected on the same grounds.
Regarding Claim 14,
Claim 14 is the non-transitory storage medium corresponding to the method of claim 4. Claim 14 is substantially similar to claim 4 and is rejected on the same grounds.
Regarding Claim 16,
Claim 16 is the non-transitory storage medium corresponding to the method of claim 6. Claim 16 is substantially similar to claim 6 and is rejected on the same grounds.
Regarding Claim 17,
Claim 17 is the non-transitory storage medium corresponding to the method of claim 7. Claim 17 is substantially similar to claim 7 and is rejected on the same grounds.
Regarding Claim 18,
Claim 18 is the non-transitory storage medium corresponding to the method of claim 8. Claim 18 is substantially similar to claim 8 and is rejected on the same grounds.
Regarding Claim 19,
Claim 19 is the non-transitory storage medium corresponding to the method of claim 9. Claim 19 is substantially similar to claim 9 and is rejected on the same grounds.
Claim 2 and 12 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Subramanya/Donovan/Dzierzanowski/Bhargava, as applied above, and further in view of Nevanperä et al. (“Aspects to Responsible Artificial Intelligence Ethics of Artificial Intelligence and Ethical Guidelines in SHAPES Project”).
Regarding Claim 2,
Subramanya, Donovan, Dzierzanowski, Bhargava teach the method as recited in claim 1.
Subramanya, Donovan, Dzierzanowski, Bhargava do not explicitly disclose
wherein the Al ethics pillars comprise accountability, value alignment, explainability, fairness, and user data rights.
However, Nevanperä (“Aspects to Responsible Artificial IntelligenceEthics of Artificial Intelligence and Ethical Guidelines in SHAPES Project”) teaches
wherein the Al ethics pillars comprise accountability, value alignment, explainability, fairness, and user data rights (pg. 35, section 5.3.1; “Firstly, they give five areas of ethical focus: Accountability, value alignment, explainability, fairness and user data rights. These can be seen as a selection of European Commission’s most important guidelines that are more straightforward to take into action.”).
Subramanya, Donovan, Dzierzanowski, Bhargava, and Nevanperä are analogous because they are directed to the same field of endeavor of ethical artificial intelligence.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the ethical AI model of Subramanya with the ethical guidelines of Nevanperä.
Doing so would allow for setting ethical guidelines for AI that takes into account different social contexts (Nevanperä pg. 20, section 5.1.3;)
Regarding Claim 12,
Claim 12 is the non-transitory storage medium corresponding to the method of claim 2. Claim 12 is substantially similar to claim 2 and is rejected on the same grounds.
Claims 10 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Subramanya/Donovan/Dzierzanowski/Bhargava, as applied above, and further in view of Kursun et al. (US-20190311277-A1).
Regarding Claim 10,
Subramanya, Donovan, Dzierzanowski, and Bhargava teach the method as recited in claim 1.
Subramanya, Donovan, Dzierzanowski, and Bhargava do not explicitly disclose
further comprising generating a historical lineage of Al ethics compliance for the asset.
However, Kursun (US 20190311277 A1) teaches
further comprising generating a historical lineage of Al ethics compliance for the asset (para [0056] “In other embodiments, the engines may be organized in an ensemble fashion. In this way each engine and its historical performance in fairness, compliance identification, and identifying other engine faults are viewed together for consideration. This ensemble system then outputs a result that is facilitating an optimization for the overall object functions. In this way, the systems stores historical performance of the individual engines in compliance violations, fairness criteria, accuracy, evaluation of other results, audits, and the like.”).
Subramanya, Donovan, Dzierzanowski, Bhargava, and Kursun are analogous because they are directed to the same field of endeavor of ethical artificial intelligence.
It would have been obvious to one of ordinary skill in the art before the effective filing date to modify the ethical AI model of Subramanya with the performance history of Kursun.
Doing so would allow for optimizing the asset based on historical performance, compliance issues, adversarial testing, audit records, accuracy, and efficiency (Kursun para [0055]).
Regarding Claim 20,
Claim 20 is the non-transitory storage medium corresponding to the method of claim 10. Claim 20 is substantially similar to claim 10 and is rejected on the same grounds.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to HENRY K NGUYEN whose telephone number is (571)272-0217. The examiner can normally be reached Mon - Fri 7:00am-4:30pm.
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, Li B Zhen can be reached at 5712723768. 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.
/HENRY NGUYEN/Examiner, Art Unit 2121
Prosecution Insights