DETAILED ACTION
This Office Action is in response to the remarks submitted on 02/19/2026. Claims 1-6, 12-17, 19 are amended. Claims 1-20 are pending.
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 Arguments
Applicant’s arguments filed on 02/19/2026 have been fully considered.
In reference to Applicant’s arguments:
- Claim rejections under 35 USC 101.
Examiner’s response:
Rejections are withdrawn in view of amendments and applicant’s arguments.
In reference to Applicant’s arguments:
- Claim rejections under 35 USC 103.
Examiner’s response:
Applicant’s arguments are mainly directed to the amended limitations and to the combination of Brenner and Aviyam failing to teach performing a specific response refinement required by the amended claims (i.e., removing unsupported portions and replacing them with context-supported information).
Examiner respectfully disagrees, as prior art Aviyam does provide an updated or modified answer to the user based on a detected wrong answer (interpreted as an hallucination). Aviyam recites, as it can be seen at [0013]: “determining a combined score according to at least the first, second, and third relevance values, evaluating the combined score against a predetermined threshold or other criteria, and determining whether to present the generated answer to the end-user according to the evaluating”, and at [0067]: “whether the answer is related to the website content and whether the website content is related to the question. Each check may be graded to determine if the answer is correct or not. This may overcome any hallucinogenic responses returned by the answer engine”. These teachings by Aviyam are not patentable distinct (emphasis added) from the amended limitations as Aviyam shows the capability of detecting wrong answers (hallucinations) according to a combined score in order to avoid hallucinogenic responses returned by the answer engine to the user. Furthermore, Aviyam recites at [0134]: “It will be appreciated that if the relevance score is low, answer engine 140 may notify the site owner that there is content missing or instruct prompt generator 141 to prompt LLM 300 to provide an updated answer to the user”; therefore, this is reasonably interpreted as the amended claim limitation “refining the response”, as an updated answer is similar to a refined answer/response.
In regards to the newly added limitation “improving the Gen AI model based on one or more actions taken to refine the response”, these arguments have been fully considered, but are moot in view of new grounds of rejection.
For these reasons, rejections are still maintained.
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 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.
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:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or nonobviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-2, 4-7, 12-13, 15-19 are rejected under 35 U.S.C. 103 as being unpatentable over Brenner et al (US Pub. 2025/0165463- hereinafter Brenner) in view of Aviyam et al (US Pub. No. 2025/0252124 - hereinafter Aviyam) and further in view of Wang et al (US Pub. No. 2025/0094866- hereinafter Wang).
Referring to Claim 1, Brenner teaches a method comprising:
providing a query and context associated with the query to a generative artificial intelligence (Gen Al) model, the Gen Al model trained to generate a response to the query based on the context (see Brenner at [0048]: “the retrieved document chunks 308 provided to the generative model 316 are said to form at least part of a context 322 associated with a user query 318, and both the user query 318 and the context 322 can be provided as inputs to the generative model 316. Collectively, the user query 318 and the context 322 can be said to form at least part of a prompt for the generative model 316. This approach helps the generative model 316 to ground the answers contained in its outputs 320 in order to mitigate issues like hallucinations and catastrophic forgetting”. Therefore, Brenner’s user query and context being inputted to the generative model being trained to output an answer is analogous to the claimed limitation);
performing analysis of the response based on a first relevancy between the query and the context, a second relevancy between the query and the response, and a third relevancy between the response and the context (see Brenner at [0031]: “Rewards can be determined based on the responses, and the RAG system can be trained using training data and augmented training data over time to optimize the RAG system (including the retriever model 112)”. Therefore, augmented training is performed base don the response to determine a reward, which is interpreted as performing analysis of the response. Further at [0059]: “the architecture 300 can incorporate support for few-shot learning examples into its structure, which can involve the configuration function 414 providing one or several examples of how the generative model 316 should operate. As a particular example, the configuration function 414 may provide to the generative model 316 (i) at least one example user query-context pair and (ii) a desired output to be generated by the generative model 316 for each example user query-context pair. Among other things, this can be useful for user queries 318 that demand precise responses or additional instances from a user”. Therefore, Brenner’s configuration function establishing a desired response according to each query-context pair is interpreted as a triadic relationship between the query , the context, and the output. Further, at [0086]: “In this data, an input prompt asks the generative model 316 to evaluate the quality of a given response to a particular instruction, and the provided evaluation result response may include a chain-of-thought reasoning (a justification) followed by a final score (such as out of a maximum score of five). Example response evaluation criteria may include relevance, coverage, usefulness, clarity, and expertise. This data is referred to as evaluation fine-tuning (EFT) data”).
However, Brenner fails to explicitly teach specifically a first, second, and third relevancies in the limitation “performing analysis of the Gen Al model based on a first relevancy between the query and the context, a second relevancy between the query and the response, and a third relevancy between the response and the context”;
identifying hallucinations and one or more sources of the hallucinations in the response based on the analysis;
refining the response based on the analysis one or more sources of the hallucinations, the response refined to address deficiencies in the query, the context, or the response associated with the hallucinations; and
improving the Gen Al model based on one or more actions taken to refine the response.
Aviyam specifically teaches a first, second, and third relevancies in the limitation “performing analysis of the Gen Al model based on a first relevancy between the query and the context, a second relevancy between the query and the response, and a third relevancy between the response and the context” (see Aviyam at [0013]: “The input processor is configured to receive the query from the end-user. The content engine is configured to retrieve content from a content management system (CMS) of the WBS relevant to the query. The prompt generator is configured to prompt the LLM to generate an answer according to the query and the content. The chat triad is configured to evaluate relevance and accuracy of the generated answer by assigning a first relevance value to a relationship between the query and the content, assigning a second relevance value to a relationship between the query and the generated answer, assigning a third relevance value to a relationship between the content and the answer, determining a combined score according to at least the first, second, and third relevance values, evaluating the combined score against a predetermined threshold or other criteria, and determining whether to present the generated answer to the end-user according to the evaluating”; wherein the query in Aviyam’s teachings corresponds to the claimed query, the content in Aviyam’s teachings corresponds to the claimed context, and the answer in Aviyam’s teachings corresponds to the claimed response);
identifying hallucinations and one or more sources of the hallucinations in the response based on the analysis (see Aviyam at [0013]: “The chat triad is configured to evaluate relevance and accuracy of the generated answer by assigning a first relevance value to a relationship between the query and the content, assigning a second relevance value to a relationship between the query and the generated answer, assigning a third relevance value to a relationship between the content and the answer, determining a combined score according to at least the first, second, and third relevance values, evaluating the combined score against a predetermined threshold or other criteria, and determining whether to present the generated answer to the end-user according to the evaluating”. In addition, per [0067]: “whether the answer is related to the website content and whether the website content is related to the question. Each check may be graded to determine if the answer is correct or not. This may overcome any hallucinogenic responses returned by the answer engine”. Therefore, Aviyam’s chat triad evaluates the scores according to certain criteria or threshold in order to determine to present the answer or not being determined as inaccurate/hallucination, meaning if the answer is correct or not (hallucination) based on the error in the three relevance values, interpreted as the sources);
refining the response based on the one or more sources of the hallucinations, the response refined to address deficiencies in the query, the context, or the response associated with the hallucinations (see Aviyam at [0134]: “It will be appreciated that if the relevance score is low, answer engine 140 may notify the site owner that there is content missing or instruct prompt generator 141 to prompt LLM 300 to provide an updated answer to the user that his request cannot be fulfilled or that an alternative solution is offered to something that might be out of stock”. Therefore, Aviyam’s determination of updating the answer to the user based on a low score (hallucination) is interpreted as refining the response based on the one or more sources).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Brenner with the above teachings of Aviyam by using a Gen AI model trained to generate a response to a user’s query, as taught by Brenner, and considering a relevancy between the query, the context of the query, and the answer in order to identify hallucinations, as taught by Aviyam. The modification would have been obvious because one of ordinary skill in the art would be motivated to evaluate relevance, accuracy, and coherence of responses generated by the Gen AI model to create contextually appropriate and accurate responses (as suggested by Aviyam at [0032]: “a chat triad configured to evaluate relevance, accuracy, and coherence of responses generated by the LLM” and [0120]: “As discussed herein above, answer engine 140 may function as a response generation system, leveraging the capabilities of LLM 300 and the output of content engine 130 to create contextually appropriate and accurate responses to user queries and chats. Answer engine 140 may generate responses that are not only informative and relevant but also interactive and tailored to the specific context of the website and user query”).
Wang teaches in an analogous system:
improving the Gen Al model based on one or more actions taken to refine the response (see Wang at [0027]: “Hallucination detector 120 identifies one or more assertions in the output and determines whether each assertion is true or false”. Further at [0042-0043]: “Thus, hallucination database 130 stores multiple entries, one for each false/inaccurate assertion. Each entry indicates a false assertion and one or more statements that have accurate information for that assertion. Fine-tuner 140 fine-tunes or retrains LLM 110 based on entries in hallucination database 130. The retraining causes LLM 110 to learn what content was inaccurate and what content is accurate. Thereafter, LLM 110 should produce accurate content when receiving the same or similar prompt to the one that triggered the inaccurate assertion”. Therefore, refining the LLM (interpreted as the Gen AI model) based on the hallucinations to learn from them for avoiding future hallucinations is similar to the claimed limitation).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Brenner and Aviyam with the above teachings of Wang by using a Gen AI model trained to generate a response to a user’s query considering a relevancy between the query, the context of the query, and the answer to detect hallucinations, as taught by Brenner and Aviyam, and using the detected hallucinations to refine the Gen Ai Model, as taught by Wang. The modification would have been obvious because one of ordinary skill in the art would be motivated to fine tune the Gen AI model to learn what content was inaccurate to produce accurate content when receiving future same or similar prompts (as suggested by Wang at [0043]: “The retraining causes LLM 110 to learn what content was inaccurate and what content is accurate. Thereafter, LLM 110 should produce accurate content when receiving the same or similar prompt to the one that triggered the inaccurate assertion”).
Referring to Claim 2, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, wherein the one or more sources of the hallucinations include one or more of the query, the context, or the response (see Aviyam at [0013]: “The chat triad is configured to evaluate relevance and accuracy of the generated answer by assigning a first relevance value to a relationship between the query and the content, assigning a second relevance value to a relationship between the query and the generated answer, assigning a third relevance value to a relationship between the content and the answer, determining a combined score according to at least the first, second, and third relevance values, evaluating the combined score against a predetermined threshold or other criteria, and determining whether to present the generated answer to the end-user according to the evaluating”. In addition, per [0067]: “whether the answer is related to the website content and whether the website content is related to the question. Each check may be graded to determine if the answer is correct or not. This may overcome any hallucinogenic responses returned by the answer engine”. Therefore, Aviyam’s chat triad evaluates the scores according to certain criteria or threshold in order to determine to present the answer or not being determined as inaccurate/hallucination, meaning if the answer is correct or not (hallucination) based on the error in the three relevance values, interpreted as the sources).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Brenner with the above teachings of Aviyam by using a Gen AI model trained to generate a response to a user’s query, as taught by Brenner, and considering a relevancy between the query, the context of the query, and the answer, as taught by Aviyam. The modification would have been obvious because one of ordinary skill in the art would be motivated to evaluate relevance, accuracy, and coherence of responses generated by the Gen AI model (as suggested by Aviyam at [0032]: “a chat triad configured to evaluate relevance, accuracy, and coherence of responses generated by the LLM”).
Referring to Claim 4, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, wherein the first relevancy is analyzed based on context relevancy metric (see Aviyam at [0013]: “The chat triad is configured to evaluate relevance and accuracy of the generated answer by assigning a first relevance value to a relationship between the query and the content”. Therefore, this first relevance value is interpreted as the context relevancy metric).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Brenner with the above teachings of Aviyam by using a Gen AI model trained to generate a response to a user’s query, as taught by Brenner, and considering a relevancy between the query, the context of the query, and the answer, as taught by Aviyam. The modification would have been obvious because one of ordinary skill in the art would be motivated to evaluate relevance, accuracy, and coherence of responses generated by the Gen AI model (as suggested by Aviyam at [0032]: “a chat triad configured to evaluate relevance, accuracy, and coherence of responses generated by the LLM”).
Referring to Claim 5, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, wherein the second relevancy is analyzed based on answer relevancy metric (see Aviyam at [0013]: “The chat triad is configured to evaluate relevance and accuracy of the generated answer by assigning a first relevance value to a relationship between the query and the content, assigning a second relevance value to a relationship between the query and the generated answer”. Therefore, this second relevance value is interpreted as the answer relevancy metric).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Brenner with the above teachings of Aviyam by using a Gen AI model trained to generate a response to a user’s query, as taught by Brenner, and considering a relevancy between the query, the context of the query, and the answer, as taught by Aviyam. The modification would have been obvious because one of ordinary skill in the art would be motivated to evaluate relevance, accuracy, and coherence of responses generated by the Gen AI model (as suggested by Aviyam at [0032]: “a chat triad configured to evaluate relevance, accuracy, and coherence of responses generated by the LLM”).
Referring to Claim 6, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, wherein the third relevancy is analyzed based on one or more of faithfulness metric or summarization metric (see Aviyam at [0013]: “The chat triad is configured to evaluate relevance and accuracy of the generated answer by assigning a first relevance value to a relationship between the query and the content, assigning a second relevance value to a relationship between the query and the generated answer, assigning a third relevance value to a relationship between the content and the answer”. Therefore, this third relevance value is interpreted as the faithful or summarization metric).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Brenner with the above teachings of Aviyam by using a Gen AI model trained to generate a response to a user’s query, as taught by Brenner, and considering a relevancy between the query, the context of the query, and the answer, as taught by Aviyam. The modification would have been obvious because one of ordinary skill in the art would be motivated to evaluate relevance, accuracy, and coherence of responses generated by the Gen AI model (as suggested by Aviyam at [0032]: “a chat triad configured to evaluate relevance, accuracy, and coherence of responses generated by the LLM”).
Referring to Claim 7, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, wherein the first relevancy, the second relevancy, and the third relevancy are represented using a first score, a second score, and a third score, respectively (see Aviyam at [0013]: “The chat triad is configured to evaluate relevance and accuracy of the generated answer by assigning a first relevance value to a relationship between the query and the content, assigning a second relevance value to a relationship between the query and the generated answer, assigning a third relevance value to a relationship between the content and the answer”. Therefore, these three relevance values are interpreted as the first, second and third scores).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the teachings of Brenner with the above teachings of Aviyam by using a Gen AI model trained to generate a response to a user’s query, as taught by Brenner, and considering a relevancy between the query, the context of the query, and the answer, as taught by Aviyam. The modification would have been obvious because one of ordinary skill in the art would be motivated to evaluate relevance, accuracy, and coherence of responses generated by the Gen AI model (as suggested by Aviyam at [0032]: “a chat triad configured to evaluate relevance, accuracy, and coherence of responses generated by the LLM”).
Referring to independent Claim 12 and Claim 19, they are rejected on the same basis as independent claim 1 since they are analogous claims.
Referring to dependent Claim 13, it is rejected on the same basis as dependent claim 2 since they are analogous claims.
Referring to dependent Claim 15, it is rejected on the same basis as dependent claim 4 since they are analogous claims.
Referring to dependent Claim 16, it is rejected on the same basis as dependent claim 5 since they are analogous claims.
Referring to dependent Claim 17, it is rejected on the same basis as dependent claim 6 since they are analogous claims.
Referring to dependent Claim 18, it is rejected on the same basis as dependent claim 7 since they are analogous claims.
Claims 3 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Brenner in view of Aviyam, in view of Wang, and further in view of Brin et al (US Pub. No. 2025/0322295 - hereinafter Brin).
Referring to Claim 3, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, whoever, fails to teach wherein the hallucinations are intrinsic or extrinsic.
Brin teaches, in an analogous system, wherein the hallucinations are intrinsic or extrinsic (see Brin at [0232]: “The faithfulness model 718 may be trained to evaluate whether the LLM output 702 exhibits signs of intrinsic or extrinsic hallucination, as described above”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Brenner, Aviyam and Wang with the above teachings of Brin by using a Gen AI model trained to generate a response to a user’s query considering a relevancy between the query, the context of the query, and the answer, as taught by Brenner, Aviyam and Wang, and detecting intrinsic or extrinsic hallucinations in the answer, as taught by Brin. The modification would have been obvious because one of ordinary skill in the art would be motivated to evaluate faithfulness of the GEN AI model for its adherence to the content within its training data (as suggested by Brin at [0153]: “Faithfulness evaluates an LLM for its adherence to the content within its training data. Low levels of faithfulness may indicate the presence of “hallucinations,” which generally refers to incorrect, nonexistent, or nonsensical output. Two types of hallucinations are generally recognized: intrinsic hallucinations where the LLM manipulates information within the input text and produces an output that is not factual with respect to the input, and extrinsic hallucinations where the LLM appears to output information that has little or no basis in the input data whatsoever”).
Referring to dependent Claim 14, it is rejected on the same basis as dependent claim 3 since they are analogous claims.
Claims 8-9 are rejected under 35 U.S.C. 103 as being unpatentable over Brenner in view of Aviyam, in view of Wang, and further in view of Xu et al (NPL “SC-Safety: A Multi-round Open-ended Question Adversarial Safety Benchmark for Large Language Models in Chinese” - hereinafter Xu).
Referring to Claim 8, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, however, fails to teach further comprising:
obtaining one or more safety policies;
assigning a safety score to the response based on one or more safety policies; and
generating a report including at least the safety score.
Xu teaches, in an analogous system,
obtaining one or more safety policies (Xu teaches at p. 3, section 2 SuperCLUE-Safety, basic moral and legal standards for testing LLms for compliance, such as Privacy and Property, Illegal Criminal Activities, Injury, and Moral Ethics. Furthermore, see Xu at p. 7, section 5 Conclusion: “The introduction of SuperCLUE-Safety(SC-Safety) provides a comprehensive and challenging benchmark for evaluating the safety of Chinese language models”. Therefore this corresponds as safety policies);
assigning a safety score to the response based on one or more safety policies (see Xu at p. 4, right column, second full paragraph: “Subsequently, based on the acquired responses, we evaluate the safety risks of the questions and follow-ups, including calculating a safety score and providing commentary. We then determine whether to include them in our final safety dataset based on the score’s indication of safety risk (1 for significant safety issues; 2 for safety issues; 3 for no safety issues). If the response presents a safety issue (scored 1 or 2), it is added to our dataset”); and
generating a report including at least the safety score (see Xu at p. 4, right column, second full paragraph: “Subsequently, based on the acquired responses, we evaluate the safety risks of the questions and follow-ups, including calculating a safety score and providing commentary. We then determine whether to include them in our final safety dataset based on the score’s indication of safety risk (1 for significant safety issues; 2 for safety issues; 3 for no safety issues). If the response presents a safety issue (scored 1 or 2), it is added to our dataset”. Therefore, the commentary corresponds to the report).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Brenner, Aviyam and Wang with the above teachings of Xu by using a Gen AI model trained to generate a response to a user’s query considering a relevancy between the query, the context of the query, and the answer, as taught by Brenner, Aviyam and Wang, and determining a safety score of the responses by the model, as taught by Xu. The modification would have been obvious because one of ordinary skill in the art would be motivated to provide a comprehensive and challenging benchmark for evaluating the safety of models, thereby promoting collaborative efforts to create safer and more trustworthy models (as suggested by Xu at Abstract: “By introducing SCSafety, we aim to promote collaborative efforts to create safer and more trustworthy LLMs” and Conclusion: “The introduction of SuperCLUE-Safety(SC-Safety) provides a comprehensive and challenging benchmark for evaluating the safety of Chinese language models”).
Referring to Claim 9, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, however, fails to teach further comprising:
assigning a safety score to the response based on the first relevancy, the second relevancy, and the third relevancy; and generating a report including at least the safety score.
Xu teaches, in an analogous system, assigning a safety score to the response based on the first relevancy, the second relevancy, and the third relevancy (see Xu at p. 4, right column, second full paragraph: “Subsequently, based on the acquired responses, we evaluate the safety risks of the questions and follow-ups, including calculating a safety score and providing commentary. We then determine whether to include them in our final safety dataset based on the score’s indication of safety risk (1 for significant safety issues; 2 for safety issues; 3 for no safety issues). If the response presents a safety issue (scored 1 or 2), it is added to our dataset”). Therefore, the safety verification system of Xu for determining if a response is safe or not would have been obvious to combine it with the relevancies of the analysis taught by Aviyam to further train the system to create safer and more trustworthy LLMs); and
generating a report including at least the safety score (see Xu at p. 4, right column, second full paragraph: “Subsequently, based on the acquired responses, we evaluate the safety risks of the questions and follow-ups, including calculating a safety score and providing commentary. Therefore, the commentary corresponds to the report).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Brenner, Aviyam and Wang with the above teachings of Xu by using a Gen AI model trained to generate a response to a user’s query considering a relevancy between the query, the context of the query, and the answer, as taught by Brenner, Aviyam and Wang, and determining a safety score of the responses by the model, as taught by Xu. The modification would have been obvious because one of ordinary skill in the art would be motivated to provide a comprehensive and challenging benchmark for evaluating the safety of models, thereby promoting collaborative efforts to create safer and more trustworthy models (as suggested by Xu at Abstract: “By introducing SCSafety, we aim to promote collaborative efforts to create safer and more trustworthy LLMs” and Conclusion: “The introduction of SuperCLUE-Safety(SC-Safety) provides a comprehensive and challenging benchmark for evaluating the safety of Chinese language models”).
Claims 10-11 and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Brenner in view of Aviyam, in view of Wang, and further in view of Ahmed et al (US Pub. No. 2025/0055867 - hereinafter Ahmed).
Referring to Claim 10, the combination of Brenner, Aviyam and Wang teaches the method of claim 1, however, fails to teach wherein the analysis includes personal identifiable information (PII) detection.
Ahmed teaches, in an analogous system, wherein the analysis includes personal identifiable information (PII) detection (see Ahmed at [0056]: “The system 102 dynamically selects the appropriate nano classifiers based on the detected threat types. For example, if a macro classifier identifies a potential PII threat, the system 102 may select nano classifiers trained to detect various subtypes of PII threats, such as social security numbers, email addresses, or phone numbers”. Further at [0094]: “Consider a scenario where a user 310 provides an input (for example, a prompt and content 312) to the generative AI model. The input may be further processed to the input sub-layer 304 of the memory 204”, and “As already explained in FIG. 2, for example, the assessment engine 208 may determine an intent and category of the prompt and content 312 based on the attributes of the data to detect threats associated with the prompt and content 312. The threats may include a prompt injection check, a jailbreak check, a profanity and toxicity check, a Personal Identifiable Information (PII) check, an Intellectual Property (IP) violation check, and an organization policy and a role-based check”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Brenner, Aviyam and Wang with the above teachings of Ahmed by using a Gen AI model trained to generate a response to a user’s query considering a relevancy between the query, the context of the query, and the answer, as taught by Brenner, Aviyam and Wang, and detecting PII in the analysis by the Gen AI model, as taught by Ahmed. The modification would have been obvious because one of ordinary skill in the art would be motivated to assess for any potential threats in the input of the Gen AI model (as suggested by Ahmed at [0094]: “The assessment engine 208 associated with the input sub-layer 304 may perform various operations on the prompt and content 312. As already explained in FIG. 2, for example, the assessment engine 208 may determine an intent and category of the prompt and content 312 based on the attributes of the data to detect threats associated with the prompt and content 312. The threats may include a prompt injection check, a jailbreak check, a profanity and toxicity check, a Personal Identifiable Information (PII) check”).
Referring to Claim 11, the combination of Brenner, Aviyam, Wang and Ahmed teaches the method of claim 10, wherein the PII detection is performed using a plurality of PII detection models (see Ahmed at [0056]: “The system 102 dynamically selects the appropriate nano classifiers based on the detected threat types. For example, if a macro classifier identifies a potential PII threat, the system 102 may select nano classifiers trained to detect various subtypes of PII threats, such as social security numbers, email addresses, or phone numbers”. These nano classifiers correspond to the plurality of PII detection models).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the combination of Brenner, Aviyam and Wang with the above teachings of Ahmed by using a Gen AI model trained to generate a response to a user’s query considering a relevancy between the query, the context of the query, and the answer, as taught by Brenner, Aviyam and Wang, and detecting PII in the analysis by the Gen AI model, as taught by Ahmed. The modification would have been obvious because one of ordinary skill in the art would be motivated to assess for any potential threats in the input of the Gen AI model (as suggested by Ahmed at [0094]: “The assessment engine 208 associated with the input sub-layer 304 may perform various operations on the prompt and content 312. As already explained in FIG. 2, for example, the assessment engine 208 may determine an intent and category of the prompt and content 312 based on the attributes of the data to detect threats associated with the prompt and content 312. The threats may include a prompt injection check, a jailbreak check, a profanity and toxicity check, a Personal Identifiable Information (PII) check”).
Referring to dependent Claim 20, it is rejected on the same basis as dependent claim 10 since they are analogous claims.
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 LUIS A SITIRICHE whose telephone number is (571)270-1316. The examiner can normally be reached M-F 9am-6pm.
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, David Yi can be reached at (571) 270-7519. 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.
/LUIS A SITIRICHE/Primary Examiner, Art Unit 2126
Prosecution Insights