DETAILED ACTION
Receipt of Applicant’s Amendment, filed May 18, 2026 is acknowledged.
Claims 21, 22, 24, 26, 28, 30, 32, 34-39 were amended.
Claims 1-20, 23, 29, 33, and 40 were cancelled.
Claims 41-45 were newly added.
Claims 21, 22, 24-28, 30, 32, 34-39, 41-45 are pending in this office 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 .
Claim Interpretation
With regard to claim 21, the claim recites “executing, by a generative artificial intelligence (Al) model, a plurality of Al prompts causing generation of prompt responses.” This claim limitation appears to recite an intended use of the claimed executing functionality. Method claims are limited by the functions that are claimed to be performed. In this limitation, the claimed function is “executing… a plurality of AI prompts” which is performed by the ‘generative artificial intelligence (AI) model”. The claimed method does not require the generation of prompt responses. This language is recited as an intended ‘cause’ resulting from the claimed execution.
It is suggested that the claims be amended to directly claim the functionality to which patent coverage is desired. To be clear, it is suggested that the claims be amended to recite the method receiving an AI prompt, and generating a response should the applicant desire such functions to be within the scope of the claims.
With regard to claim 21, the claim recites “generative artificial intelligence (AI) model” and “machine learning model”. For examination purposes these claim limitations have been interpreted in light of Figure 1, as referring to two distinct claim elements.
With regard to claim 24 “storing the development prompt record for access by the AI development system.” This claim limitation is a recitation of intended use of the claimed storage operation. What the data is stored for access for does not impose a functional limitation on the storage of the data.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 21, 22, 24-28, 30, 32, 34-39, 41-45 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
With regard to claim 21, the claim recites “receiving a generative artificial intelligence (Al) model application programming interface (API): in response to receiving the generative Al model API call, dynamically select an Al prompt among the plurality of Al prompts stored within a prompt record and a response generated by a generative Al model based at least in part on the Al prompt” This claim limitation lacks antecedent basis.
One of ordinary skilled in the art would recognize that a received API is referring to an API call. This raises questions regarding the distinction between the recited “generative artificial intelligence (AI) model application program interface (API)” that is received and the recited “generative AI model API call”.
Furthermore, the instant claim language appears to recite a “:” after the recited receiving of the API. This symbol indicates that the following stanza is included in the instant stanza. One of ordinary skill in the art may reasonably read this claim language to mean that receiving the API call is in response to the API call, which is logically inconsistent and renders the meaning of the claim unclear.
It is unclear what the “response” is part of within the claim language. The grammar of the claim renders the meaning of the claim unclear.
The claim has previously recited “a generative AI model”. It is unclear if applicant is reciting a new generative AI model or attempting to refer to the previously recited generative AI model.
For examination purposes this claim limitation has been construed to mean -- “receiving a generative artificial intelligence (Al) model application programming interface (API) call; in response to receiving the API call, dynamically select an Al prompt and a response; wherein the selected AI prompt is selected from among the plurality of Al prompts stored within a prompt record; and wherein the response is generated by the generative Al model based at least in part on the selected Al prompt –
Claim 39 recites substantially similar language and is rejected based upon the same reasoning and rational.
With regard to claim 21, the claim recites “enabling access of the prompt record by exposing an API interface to client applications.” This claim limitation is confusing and unclear. It is unclear what functionality is required to be performed by the claimed method vs what is being recited as an intended use.
The claim does not recite accessing the prompt record. Instead, the claim recites that access in ‘enabled’ but not performed by the claimed method. Enabling is not a function to be performed. Within the disclosed device, the ‘enabling’ of access being achieved by “exposing an API interface to client applications”. One of ordinary skill in the art would recognize this to mean that an API is present. The presence of an API is not a function to be performed by the claimed method. It is suggested that the claims be amended to recite the functionality that is performed.
Furthermore, it should be noted that the claim has already recited receiving an API call, and in response to said API call, selecting AI prompt and response from within the prompt. One of ordinary skill in the art would recognize this as the use of the API to access the prompt record to select the AI prompt and responses. It is unclear if applicant is attempting to recite a second access to the prompt record, or if applicant is referring to the previously recited functionality of selecting data from the prompt record in response to the API call.
For examination purposes this claim limitation has been construed as referring to the selection of the AI prompt and response in response to the API call.
Claim 39 recites substantially similar language and is rejected based upon the same reasoning and rational.
With regard to claim 24, the claim recites “a response generated by a generative AI model”. The parent claim has already recited a response generated by a generative AI model. It is unclear if applicant is referring to the previously recited response or reciting a new response. It is unclear if applicant is referring to the previously recited generative AI model or reciting a new generative AI model.
With regard to claims 27 and 30, claim recites 27 “The computer implemented method of claim 21 wherein generating the prompt record comprises:…” This claim limitation lacks antecedent basis. Claim 30 recite similar language and is rejected based upon the same reasoning and rational.
The method of claim 21 does not perform the function of generating the prompt record. Claim 21 refers to a response generated by a generative AI model. The method of claim 21 does not perform the generation of the response. The response generation is recited as being something outside of the claimed method, wherein the response is provided to the claimed method. The recitation of how the response is generated does not invoke any functional limitation on the claimed method, as the claimed method only references to the finished product (e.g. the response that is retrieved).
It is suggested that the claimed method be amended to recite the functionality to which patent coverage is desired instead of attempting to recite the results. For examination purposes this claim limitation has been construed as if claim 21 had recited –generating, by the AI model, a response--.
With regard to claim 28, the claim recites “identifying whether the AI prompt is a surreptitious AI prompt based at least in part on an assessment of the AI prompt; … wherein processing the AI prompt to identify whether the AI prompt is the surreptitious prompt comprises:” This claim limitation lacks antecedent basis. It is unclear if the processing of the AI prompt to identify whether the AI prompt is the surreptitious prompt is the claimed assessment or if applicant is reciting a distinct processing operation. For examination purposes this claim limitation has been construed to mean – identifying whether the AI prompt is a surreptitious AI prompt based at least in part on an assessment of the AI prompt;… wherein the assessment comprises:..—
With regard to claims 30, claim 30 recites “generating the prompt record with, as the evaluation data, a set of evaluation metrics and evaluation metric values”. This claim limitation lacks antecedent basis.
The parent claim has recited “evaluation metrics” and “prompt evaluation data”. It is unclear if applicant is attempting to recite a new claim element or refer to the previously recited element. It is suggested that the claim labels be used consistently throughout the claims.
The recitation of “generating the prompt record” lacks antecedent basis. The claim limitation is written in a manner that suggests the claim is reciting a new generating step. Parent claim 21 has already recited “a prompt record” but does not recite generating said prompt record. It is unclear if applicant is attempting to refer to the previously recited prompt record, or if applicant is attempting to recite the generation of a new data element.
The recitation of “a generative AI model” lacks antecedent basis, as it is unclear if applicant is attempting to refer to the previously recited claim limitation or attempting to recite a new AI model.
Parent claim 21 already recites “prompt evaluation data indicative of a performance of the generative AI prompt”. As such, the language recited herein has been identified as duplicative language repeating the limitations already set forth in parent claim 21. The recitation of “evaluation data” recited herein lacks antecedent basis as it is unclear if applicant is attempting to recite a new claim element or attempting to refer to the previously recited element. Furthermore, the distinction between “evaluation data”, “set of evaluation metrics” and “evaluation metric values” is unclear. One of ordinary skill in the art may reasonable read all of these claim elements as referring to the same thing.
For examination purposes this claim limitation has been construed to mean – generating the prompt record with the prompt evaluation data… wherein the prompt evaluation data is generated by the generative AI model –
With regard to claim 32, the claim recites “wherein processing the AI prompt to generate a prompt record comprises: generating the prompt record with an indication of context data and augmented data used by the generative AI model in generating the response.”
With regard to the “indication” it is unclear if applicant is attempting to recite a new claim element or if applicant is attempting to refer to the “prompt content data” which is “indicative of content of the generative AI prompt”. It is noted that ‘indication’ is a noun, referring to an element, ‘indicative of’ an adjective modifying a noun. One of ordinary skill in the art may reasonably read the recited ‘indication’ as referring to the content data that was recited as being ‘indicative of’ the prompt content data. Yet the use of a distinct claim label renders is reasonable to interpret the indication as being a distinct from the prompt content data.
For examination purposes this claim limitation has been construed to mean – the prompt record indicates context data and augmented data —. Should applicant intend to recite a new claim element, it is suggested that that claim element be giving a clearly distinct claim label.
With regard to claim 34, recites “generating the prompt record with an indication of…”. This claim limitation lacks antecedent basis.
With regard to the “indication” it is unclear if applicant is attempting to recite a new claim element or if applicant is attempting to refer to the “prompt content data” which is “indicative of content of the generative AI prompt”.
For examination purposes this claim limitation has been construed to mean –generating the prompt record indicative of….-- Should applicant intend to recite a new claim element, it is suggested that that claim element be giving a clearly distinct claim label.
With regard to claim 35, the claim recites “calling an interface exposed by a generative artificial intelligence (AI) model application programming interface (API) to receive an AI prompt and a response generated by a generative AI model based at least in part on the AI prompt”. This claim limitation lacks antecedent basis.
Each unique claim label is expected to refer to a unique claim element. It is unclear how many interfaces are being recited herein. One of ordinary skill in the art would recognize an API as an interface. It is unclear if applicant is referring to the API or attempting to recite a distinct interface.
The claimed functionality performed by the claimed system is calling the API. The results of that call is ‘to receive’ the AI prompt and response. The claimed system is not recited as performing the function of receiving the AI prompt and response, or generating said AI prompt or response. The receiving of the AI prompt and response has been read as an intended use of the claimed API call. The generating of the AI prompt and response is recited as being performed by a separate device that is not part of the claimed device itself.
It is suggested that the claim be amended to recite the structure of the claimed system and the functionality performed by the claimed system instead of reciting intended uses and elements external to the claimed system.
For examination purposes this claim limitation has been construed to mean – calling an API; receiving, in response to the API call, an AI prompt and response--.
Claim Objections
Claims 21, 22, 24-28, 30, 32, 34-38, 41-45 are objected to because of the following informalities. Appropriate correction is required.
With regard to claim 21, the claim recites “the prompt record including prompt content data indicative of content of the generative Al prompt”. This claim limitation appears to add needless linguistics that do not further limit the claimed device, thereby obfuscating the claimed device.
The label plain language of “prompt content data” constitutes content data that is a prompt. Therefore, the language that the prompt content data is “data indicative of content of the generative Al prompt” does not appear to add any further limitation or clarity of what the “prompt content data” is. Instead, it obfuscates the meaning of the claim as it suggests that the prompt content data is merely indicative of content of the prompt instead of being the prompt content data.
It is suggested that duplicative language that suggests broader interpretations than the plain meaning of the claim element be removed.
Claim 22 recites similar language (e.g. “the response record including response content data indicative of content of the response”) that is objected to for the same reasoning and rational and has been interpreted similarly.
Claim 35 recites similar language (e.g. “the prompt record including prompt content data indicative of content of the generative AI prompt”) that is objected to for the same reasoning and rational and has been interpreted similarly.
With regard to claim 38, the claim recites “interacting with an AI development system to receive a development system AI prompt and a response generated by a generative AI model based at least in part on the development system AI prompt;”
The claimed operation is ‘interacting’ the result of the interacting is ‘to receive. The claimed system is not the AI development system. The AI development system appears to be external to the claimed system. The operations of the AI development system are external and do not invoke a functional limitation on the claimed system. The claim appears to recite the intention to use this external AI developmental system to receive the developmental AI prompt and response. The claimed system does not recite receiving said developmental AI prompt and response. But instead recites the intention the receive said data in response an undefined ‘interaction’.
The claim is formulated in a manner in which it is difficult to determine what is part of the claimed device, and what operations are performed by the claimed device. As detailed above, the claimed functionality appears to be performed by external system, and not within the scope of the claimed device itself. This renders the scope of the claimed device unclear.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(1) the claimed invention was patented, described in a printed publication, or in public use, on sale, or otherwise available to the public before the effective filing date of the claimed invention.
Claims 21, 22, 24-27, 30, 32, 34-36, 38, 39, 41, and 45 are rejected under 35 U.S.C. 102(a)(1) as being anticipated by Safronov904 [2020/0192904].
With regard to claim 21 Safronov904 teaches A computer implemented method, comprising:
executing, by a generative artificial intelligence (Al) model as a first MLA 126 of the search server 120 (Safronov904, ¶86; ¶99 “the search engine server 120 executes one or more machine learning algorithms (MLAs) 126”), a plurality of Al prompts as given search queries (Safronov904, ¶88 “(ii) execute searches in response to a given search query”) causing generation of prompt responses as search results (Safronov904, ¶88 “(iii) execute analysis of documents and perform ranking of documents in response to the given search query; (iv) group the documents and compile the search engine result page (SERP) to be outputted to a client device (such as one of the first client device 104, the second client device 106, the third client device 108 and the fourth client device 110), the client device having been used to submit the given search query that resulted in the SERP.”);
assessing, by a machine learning model as a second MLA 126 (Safronov904, ¶99 “the search engine server 120 executes one or more machine learning algorithms (MLAs) 126”), evaluation metrics as retrieving respective values 334 of a meta-feature 282 (Safronov904, ¶173 “The search engine server 120 may query the training database 142 to verify if the current query 304 is one of the set of past queries 202, in which case the search engine server 120 also retrieves respective values 334 of a meta-feature 282 computed for the respective set of past documents (not depicted) associated with the respective past query (not depicted) similar to the current query 304, which will be used for ranking the set of current documents 310, the set of current documents 310 including documents in the set of past documents.”) based at least in part on a comparison as the similarity (Id) between the plurality of Al prompts as comparing the past queries 202 to the current query 304 (Id) and the prompt responses as s the respective set of past documents associated with the past queries (Id);
receiving a generative artificial intelligence (Al) model application programming interface (API) (Safronov904, ¶56 “appropriate hardware and is capable of receiving requests (e.g. from electronic devices) over a network, and carrying out those requests, or causing those requests to be carried out.”):
in response to receiving the generative Al model API call (Id; Please see the 112b above, this claim limitation has been construed as referring to the API call received), dynamically select an Al prompt as the determined similar past query (Safronov904, ¶173) among the plurality of Al prompts as the past query (Safronov904, ¶173) stored within a prompt record as query log 136 (Safronov904, ¶117 “the query log 136 may include a list of queries with their respective terms, with information about documents that were listed by the search engine server 120 in response to a respective query, a timestamp, and may also contain a list of users identified by anonymous IDs ( or without an ID altogether) and the respective documents they have clicked on after submitting a query”) and a response as the respective set of past documents associated with the past queries (Id) generated by [a generative Al model] as a first MLA 126 of the search server 120 (Safronov904, ¶86; ¶99; Please see the 112b above, this claim limitation has been construed to mean –the AI model) based at least in part on the Al prompt as the current query that is used to select the past similar queries (Safronov904, ¶173), the selection being based at least in part on the evaluation metrics as retrieving respective values 334 of a meta-feature 282 (Safronov904, ¶173), the prompt record as query log 136 (Safronov904, ¶117) including prompt content data indicative of content of the generative Al prompt as the associated search words of the search query (Safronov904, ¶116 “the query log 136 maintains terms of search queries (i.e. the associated search words) and the associated search results.”) and prompt evaluation data indicative of a performance of the generative Al prompt (Safronov904, ¶146 “As a non-limiting example, the first plurality of features 220 may include indications of user interactions or user engagement metrics tracked and compiled by the tracking server 130 such as one or more of:” ¶147-¶150); and
[enabling access of the prompt record as query log 136 (Safronov904, ¶117) by exposing an API interface to client applications] (Safronov904, ¶56 “appropriate hardware and is capable of receiving requests (e.g. from electronic devices) over a network, and carrying out those requests, or causing those requests to be carried out.”; Please see the 112b above, this claim limitation has been construed as referring to the dynamically selected AI prompt and response).
With regard to claim 22 Safronov904 further teaches
generating a response record based at least in part on the response received from the generative Al model API, the response record including response content data indicative of content of the response as user interaction log storing the reference document ID (Safronov904, ¶118 “As a non-limiting example, the user interaction log 138 may contain a reference to a document, which may be identified by an ID number or an URL, a list of queries, where each query of the list of queries has been used to access the document, and respective user interactions associated with the document for the respective query of the list of queries (if the document was interacted with)”); and
storing the response record as the user interaction log (Id) with a record indicator indicating that the response record is related to the prompt record as the reference document ID and the list of queries (Id), wherein processing the Al prompt to generate the response record as the user interaction log (Id) comprises generating the response record to include user interaction indicators indicative of user interactions as user interactions (Id) with the response as associated with the document (Id).
With regard to claim 24 Safronov907 further teaches
obtaining, from an AI development system as the training server (Safronov904, ¶127 “The training server 140 acquires a set of past queries 202 from the query log 136”), a development system AI prompt as one of the past queries 202, e.g. first past query 204 (Safronov904, ¶127 “The training server 140 acquires a set of past queries 202 from the query log 136”; ¶130) and [a response as the past document 140 (Safronov904, ¶130 “The training server 140 acquires, for the first past query 204, a set of past documents 210, the set of past documents 210 having been presented as search results in a search engine results page (SERP) to one or more of the plurality of client devices 102 in response to the first past query 204 having been submitted on the search engine server 120”) generated by a generative AI model] as the past documents determined by the MLA126 (Safronov904, ¶131 “The set of past documents 210 generally includes a predetermined number of documents, such as the top 100 most relevant documents that have been presented in a SERP in response to the first past query 204, as determined by the MLA 126 of the search engine server 120.”; Please note this claim limitation has been construed to mean –the response generated by the generative AI model--) based at least in part on the development system AI prompt as the first past query 204 (Id);
generating a development prompt record as the training database 142 (Safronov904, ¶124) based at least in part on the development system AI prompt received from the Al development system as selecting the past queries (Safronov904, ¶128 “How the training server 140 selects queries to be part of the set of past queries 202 is not limited”), the development prompt record including prompt content data indicative of content of the development system AI prompt (Safronov904, ¶117 “More specifically, the query log 136 may include a list of queries with their respective terms”) and prompt evaluation data indicative of a performance of the development system AI prompt (Safronov904, ¶146 “As a non-limiting example, the first plurality of features 220 may include indications of user interactions or user engagement metrics tracked and compiled by the tracking server 130 such as one or more of:” ¶147-¶150); and
storing the development prompt record for access by the AI development system (Safronov904, ¶168 “In other embodiments, the value 284 of the meta-feature 282 for each respective document 212 of the set of past documents 210 may be stored together with the first plurality of features 220 in the index 124 and/or the query log 136 and/or the user interaction log 138.”).
With regard to claim 25 Safronov904 further teaches
receiving a retrieval call from the generative AI model API as quiring the training database 142 (¶173, “The search engine server 120 may query the training database 142 to verify if the current query 304 is one of the set of past queries 202, in which case the search engine server 120 also retrieves respective values 334 of a meta-feature 282 computed for the respective set of past documents (not depicted) associated with the respective past query (not depicted) similar to the current query 304, which will be used for ranking the set of current documents 310, the set of current documents 310 including documents in the set of past documents”); and
returning a prompt record to the generative AI model API based at least in part on the retrieval call as retrieving the respective values (Id).
With regard to claim 26 Safronov904 further teaches automatically populating a prompt library in a user data storage system with the prompt record as the search log database 122 (Safronov904, ¶119 “which may store the tracked queries, user interactions and associated search results in the search log database 122.”).
With regard to claim 27 Safronov904 further teaches wherein generating the prompt record comprises:
identifying tokens in the AI prompt as search words (¶166 “More specifically, the query log 136 maintains terms of search queries (i.e. the associated search words) and the associated search results.”); and
populating the prompt record with an indication of the tokens in the AI prompt as the query log 136 maintains the terms of the search query (Id).
With regard to claim 30 Safronov904 further teaches wherein generating the prompt record comprises:
generating the prompt record as query log 136 (Safronov904, ¶117) with, as evaluation data as retrieving respective values 334 of a meta-feature 282 (Safronov904, ¶173 “The search engine server 120 may query the training database 142 to verify if the current query 304 is one of the set of past queries 202, in which case the search engine server 120 also retrieves respective values 334 of a meta-feature 282 computed for the respective set of past documents (not depicted) associated with the respective past query (not depicted) similar to the current query 304, which will be used for ranking the set of current documents 310, the set of current documents 310 including documents in the set of past documents.”; Please see the 112b above, this claim limitation has been construed as referring to the evaluation data recite din claim 21), a set of evaluation metrics as the meta0features 282 (Id) and evaluation metric values as the respective values (Id) indicative of the performance of the generative AI prompt(Safronov904, ¶146 “As a non-limiting example, the first plurality of features 220 may include indications of user interactions or user engagement metrics tracked and compiled by the tracking server 130 such as one or more of:” ¶147-¶150),
[wherein generating the prompt record as query log 136 (Safronov904, ¶117) with, as the evaluation data, the set of evaluation metrics and evaluation metric values (Safronov904, ¶173) comprises receiving the set of evaluation metrics and metric values from a generative AI evaluation model] (¶107 “Generally speaking, the tracking server 130 is configured to track user interactions with search results provided by the search engine server 120 in response to user requests ( e.g. made by users of one of the first client device 104, the second client device 106, the third client device 108 and the fourth client device 110). The tracking server 130 may track user interactions (such as, for example, clickthrough data) when users perform general web searches and vertical web searches on the search engine server 120, and store the user interactions in a tracking database 132.”; Please see the 112b above, this claim limitation has been construed to mean that --the prompt evaluation data is generated by the generative AI model--).
With regard to claim 32 Safronov904 further teaches wherein processing the AI prompt to generate the prompt record comprises:
generating the prompt record (Safronov904, ¶119 “In some embodiments, the tracking server 130 may send tracked queries, search result and user interactions to the search engine server 120, which may store the tracked queries, user interactions and associated search results in the search log database 122”) with [an indication of context data and augmented data] as click-through rate, number of clicks on an element, etc. (Safronov904, ¶108 “the tracking server 130 may compute a click-through rate (CTR), at predetermined intervals of time or upon receiving an indication, based on a number of clicks on an element and number of times the element was shown (impressions) in a SERP.”; Please note this claim limitation has been construed to mean –the prompt record indicates context data and augmented data--) used by the generative AI model in generating the response (¶119 “the tracking server 130 may send tracked queries, search result and user interactions to the search engine server 120, which may store the tracked queries, user interactions and associated search results in the search log database 122.”),
wherein generating the prompt record with the indication of the context data and the augmented data (¶118 “The user interaction log 138 may be linked to the query log 136, and list user interactions as tracked by the tracking server 130 after a user has submitted a query and clicked on one or more documents on a SERP on the search engine server 120.”) comprises:
identifying data extraction scripts corresponding to the prompt as document retrieval operations, e.g. search query (Safronov904, ¶172 “he search engine server 120 retrieves, from the index 124, based on terms of the current query 304”); and
generating the prompt record with an [indication] of the data extraction scripts as storing the tracked queries, user interactions and associated search results in the search log database 122 (Safronov904, ¶119 “In some embodiments, the tracking server 130 may send tracked queries, search result and user interactions to the search engine server 120, which may store the tracked queries, user interactions and associated search results in the search log database 122”; Please note this claim limitation has been construed to mean –generating the prompt record indicating the data extraction scripts--).
With regard to claim 34 Safronov904 further teaches wherein processing the AI prompt to generate the prompt record comprises:
identifying a type as the type of the evaluation metric (Safronov904, ¶188 “A type of the evaluation metric used to evaluate the usefulness of the meta-feature 282, and a type of the user interactions 378 and user interactions 388 being evaluated”) of the generative AI model as the evaluation metric used to train the training server, which is part of the MLA (¶188, ¶99) that generated the response as the set of past documents (Safronov904, ¶160 “the value 262 of the parameter 260 associated with the set of past documents 210.”);
identifying model parameters (Safronov904, ¶160 “the value 262 of the parameter 260 associated with the set of past documents 210.”) used with the identified type as the type of the evaluation metric (Safronov904, ¶188 “A type of the evaluation metric used to evaluate the usefulness of the meta-feature 282, and a type of the user interactions 378 and user interactions 388 being evaluated”) of generative AI model as the evaluation metric used to train the training server, which is part of the MLA (¶188, ¶99); and
generating the prompt record as query log 136 (Safronov904, ¶117 “the query log 136 may include a list of queries with their respective terms, with information about documents that were listed by the search engine server 120 in response to a respective query, a timestamp, and may also contain a list of users identified by anonymous IDs ( or without an ID altogether) and the respective documents they have clicked on after submitting a query”) with [an indication] (Please note this claim limitation has been construed to mean --indicative--) of the identified type as the type of the evaluation metric (Safronov904, ¶188 “A type of the evaluation metric used to evaluate the usefulness of the meta-feature 282, and a type of the user interactions 378 and user interactions 388 being evaluated”) of generative AI model as the evaluation metric used to train the training server, which is part of the MLA (¶188, ¶99) and the identified model parameters (Safronov904, ¶160 “the value 262 of the parameter 260 associated with the set of past documents 210.”) used by the generative AI model in generating the response as the set of documents 210 (Id).
With regard to claim 35 Safronov904 teaches A computer system, comprising one or more processors configured to perform operations (¶49), the operations comprising:
calling an interface as the communication link 114 (Safronov904, ¶80, “The system 100 comprises a search engine server 120, a tracking server 130 and a training server 140 coupled to the communication network 112 via their respective communication link 114”) exposed by a generative artificial intelligence (AI) model as a first MLA 126 of the search server 120 (Safronov904, ¶86; ¶99 “the search engine server 120 executes one or more machine learning algorithms (MLAs) 126”) application programming interface (API) (Safronov904, ¶56 “appropriate hardware and is capable of receiving requests (e.g. from electronic devices) over a network, and carrying out those requests, or causing those requests to be carried out.”) to receive an AI prompt as the current query (Safronov904, ¶96 “when a given query ( such as a current query of a user of the first client device 104, for example) is received by the search engine server 120”) and a response as the ranked documents (Safronov904, ¶99 “the search engine server 120 executes one or more machine learning algorithms (MLAs) 126 for ranking documents in response to the given query”) generated by a generative AI model based at least in part on the AI prompt as in response to the given query (Safronov904, ¶99 “the search engine server 120 executes one or more machine learning algorithms (MLAs) 126 for ranking documents in response to the given query”; Please see the 112b above, this claim limitation has been construed to mean – calling an API; receiving, in response to the API call, an AI prompt and response--);
generating a prompt record as query log 136 (Safronov904, ¶117 “the query log 136 may include a list of queries with their respective terms, with information about documents that were listed by the search engine server 120 in response to a respective query, a timestamp, and may also contain a list of users identified by anonymous IDs ( or without an ID altogether) and the respective documents they have clicked on after submitting a query”) based at least in part on the AI prompt as the respective quarriers (Id) received from the generative AI model API as the queries submitted via the respective client devices (Safronov904, ¶127 “acquires a set of past queries 202 from the query log 136, where each query of the set of past queries 202 has been previously submitted on the search engine server 120 by one or more users via respective associated client devices”) through the devices capable for transmitting through the network (Safronov904, ¶56), the prompt record including prompt content data indicative (Safronov904, ¶61 see definition for ‘indication’) of content as the respective terms (Safronov904, ¶117 “the query log 136 may include a list of queries with their respective terms, with information about documents that were listed by the search engine server 120 in response to a respective query, a timestamp, and may also contain a list of users identified by anonymous IDs ( or without an ID altogether) and the respective documents they have clicked on after submitting a query”) of the generative Al prompt as the respective query (Id) and prompt evaluation data indicative of a performance of the generative Al prompt as the user’s click indications (Id); and
automatically providing the prompt record for storage as storing the tracked queries, user interactions and associated search results in the search log database 122 (Safronov904, ¶119 “In some embodiments, the tracking server 130 may send tracked queries, search result and user interactions to the search engine server 120, which may store the tracked queries, user interactions and associated search results in the search log database 122”) in a data store in a user data storage system as the database 112 (Id).
With regard to claim 36 Safronov904 further teaches wherein the operations further comprise generating a response record based at least in part on the response received from the generative AI model API, the response record including response content data indicative of content of the response as user interaction log storing the reference document ID (Safronov904, ¶118 “As a non-limiting example, the user interaction log 138 may contain a reference to a document, which may be identified by an ID number or an URL, a list of queries, where each query of the list of queries has been used to access the document, and respective user interactions associated with the document for the respective query of the list of queries (if the document was interacted with)”) and user interaction indicators indicative of user interactions with the response as the respective user interactions ((Safronov904, ¶118 “As a non-limiting example, the user interaction log 138 may contain a reference to a document, which may be identified by an ID number or an URL, a list of queries, where each query of the list of queries has been used to access the document, and respective user interactions associated with the document for the respective query of the list of queries (if the document was interacted with)”).
With regard to claim 38 Safronov904 further teaches
interacting with an AI development system as the training server (Safronov904, ¶127 “The training server 140 acquires a set of past queries 202 from the query log 136”) to receive a development system AI prompt as one of the past queries 202, e.g. first past query 204 (Safronov904, ¶127 “The training server 140 acquires a set of past queries 202 from the query log 136”; ¶130) and a response as the past document 140 (Safronov904, ¶130 “The training server 140 acquires, for the first past query 204, a set of past documents 210, the set of past documents 210 having been presented as search results in a search engine results page (SERP) to one or more of the plurality of client devices 102 in response to the first past query 204 having been submitted on the search engine server 120”) generated by a generative AI model as the past documents determined by the MLA126 (Safronov904, ¶131 “The set of past documents 210 generally includes a predetermined number of documents, such as the top 100 most relevant documents that have been presented in a SERP in response to the first past query 204, as determined by the MLA 126 of the search engine server 120.”) based at least in part on the development system AI prompt as the first past query 204 (Id); and
generating a development prompt record as the training database 142 (Safronov904, ¶124) based at least in part on the development system AI prompt received from the AI development system, as selecting the past queries (Safronov904, ¶128 “How the training server 140 selects queries to be part of the set of past queries 202 is not limited”), the development prompt record including prompt content data indicative of content of the development system AI prompt (Safronov904, ¶117 “More specifically, the query log 136 may include a list of queries with their respective terms”) and prompt evaluation data indicative of a performance of the development system AI prompt (Safronov904, ¶146 “As a non-limiting example, the first plurality of features 220 may include indications of user interactions or user engagement metrics tracked and compiled by the tracking server 130 such as one or more of:” ¶147-¶150).
With regard to claim 39 Safronov904 teaches A computing system, comprising:
one or more processors as a processor (Safronov904, ¶37 “the server comprising: a processor, a non-transitory computer-readable medium comprising instructions.”); and a memory storing computer executable instructions which, when executed by the one or more processors (Id), cause the one or more processors to perform steps, comprising:
executing, by a generative artificial intelligence (Al) model accessing system as a first MLA 126 of the search server 120 (Safronov904, ¶86; ¶99 “the search engine server 120 executes one or more machine learning algorithms (MLAs) 126”), a plurality of Al prompts as given search queries (Safronov904, ¶88 “(ii) execute searches in response to a given search query”) causing generation of prompt responses as search results (Safronov904, ¶88 “(iii) execute analysis of documents and perform ranking of documents in response to the given search query; (iv) group the documents and compile the search engine result page (SERP) to be outputted to a client device (such as one of the first client device 104, the second client device 106, the third client device 108 and the fourth client device 110), the client device having been used to submit the given search query that resulted in the SERP.”), a plurality of prompt records as query log 136 (Safronov904, ¶117 “the query log 136 may include a list of queries with their respective terms, with information about documents that were listed by the search engine server 120 in response to a respective query, a timestamp, and may also contain a list of users identified by anonymous IDs ( or without an ID altogether) and the respective documents they have clicked on after submitting a query”) comprising the plurality of AI prompts as the queries (Id) and the prompt responses as the documents (Id);
assessing, by a machine learning model as a second MLA 126 (Safronov904, ¶99 “the search engine server 120 executes one or more machine learning algorithms (MLAs) 126”), evaluation metrics as retrieving respective values 334 of a meta-feature 282 (Safronov904, ¶173 “The search engine server 120 may query the training database 142 to verify if the current query 304 is one of the set of past queries 202, in which case the search engine server 120 also retrieves respective values 334 of a meta-feature 282 computed for the respective set of past documents (not depicted) associated with the respective past query (not depicted) similar to the current query 304, which will be used for ranking the set of current documents 310, the set of current documents 310 including documents in the set of past documents.”) based at least in part on a comparison as the similarity (Id) between the plurality of Al prompts as comparing the past queries 202 to the current query 304 (Id) and the prompt responses as s the respective set of past documents associated with the past queries (Id);
receiving a generative artificial intelligence (AI) model application programming interface (API) Safronov904, ¶56 “appropriate hardware and is capable of receiving requests (e.g. from electronic devices) over a network, and carrying out those requests, or causing those requests to be carried out.”):
in response to receiving the generative Al model API call (Id; Please see the 112b above, this claim limitation has been construed as referring to the API call received), dynamically selecting a prompt record as the determined similar past query (Safronov904, ¶173) of the plurality of Al prompt records as the past query (Safronov904, ¶173) based at least in part on evaluation metrics as retrieving respective values 334 of a meta-feature 282 (Safronov904, ¶173), the prompt record as query log 136 (Safronov904, ¶117 “the query log 136 may include a list of queries with their respective terms, with information about documents that were listed by the search engine server 120 in response to a respective query, a timestamp, and may also contain a list of users identified by anonymous IDs ( or without an ID altogether) and the respective documents they have clicked on after submitting a query”) including prompt content data indicative of content of the generative AI prompt as the associated search words of the search query (Safronov904, ¶116 “the query log 136 maintains terms of search queries (i.e. the associated search words) and the associated search results.”) and prompt performance data indicative of a performance of the generative AI prompt(Safronov904, ¶146 “As a non-limiting example, the first plurality of features 220 may include indications of user interactions or user engagement metrics tracked and compiled by the tracking server 130 such as one or more of:” ¶147-¶150); and
enabling access of the prompt record as query log 136 (Safronov904, ¶117) to an AI system by exposing an API interface (Safronov904, ¶56 “appropriate hardware and is capable of receiving requests (e.g. from electronic devices) over a network, and carrying out those requests, or causing those requests to be carried out.”) to the AI system (Safronov904, ¶56 “appropriate hardware and is capable of receiving requests (e.g. from electronic devices) over a network, and carrying out those requests, or causing those requests to be carried out.”; Please see the 112b above, this claim limitation has been construed as referring to the dynamically selected AI prompt and response).
With regard to claim 41 Safronov904 further teaches
capturing one or more user interaction indicators indicative of user edits modifications of the document (¶145 “Time: time based features, such as creation time of the document, modification time of the document, and the like”) to the prompt responses as the user interacting with the document (¶118 “respective user interactions associated with the document for the respective query of the list of queries (if the document was interacted with)”), wherein the evaluation metrics are generated based at least in part on the one or more user interaction indicators (Id); and
wherein the selection of the AI prompt among the plurality of AI prompts as the determined similar past query (Safronov904, ¶173) is further based at least in part on the user interaction indicators as retrieving respective values 334 of a meta-feature 282 (Safronov904, ¶173).
With regard to claim 45 Safronov904 further teaches identifying a type of a generative AI system that includes the generative AI model as the type of search (¶104 “In some embodiments of the present technology, the search engine server 120 can execute ranking for several types of searches, including but not limited to, a general search and a vertical search.”), the type of the generative AI system comprising a plurality of functionalities of the generative AI system as general or vertical search (Id);
wherein the selection of the AI prompt among the plurality of AI prompts as the determined similar past query (Safronov904, ¶173) is further based at least in part on the plurality of functionalities of the generative AI system as the search (¶104).
Claim Rejections - 35 USC § 103
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 28 and 37 are rejected under 35 U.S.C. 103 as being unpatentable over Safronov904 [20200192904] in view of Martin [20180004948].
With regard to claims 28 and 37 Safronov904 further teaches
[[
as query log 136 (Safronov904, ¶117) [[
[[ (Please note this claim limitation has been construed as referring to the claimed --assessment--):
Safronov904 does not explicitly teach identifying whether the AI prompt is a surreptitious AI prompt based at least in part on an assessment of the AI prompt; and in response to determining that the AI prompt is the surreptitious AI prompt, tagging the prompt record to identify the AI prompt as the surreptitious AI prompt, wherein processing the AI prompt to identify whether the AI prompt is the surreptitious prompt comprises: generating a prompt vector based at least in part on the AI prompt; comparing the prompt vector to a surreptitious prompt vector generated from the surreptitious prompt to obtain a comparison result; and identifying whether the AI prompt is the surreptitious prompt based at least in part on the comparison result.
Martin teaches identifying whether the AI prompt is a surreptitious AI prompt as identifying and recording a signal as at risk of being an attack signal (Martin, ¶24 “can automatically record an attempted exploitation of a vulnerability within the network in one stage of an attack as a signal, such as: presence of malware and malicious code; instances of tunneling; presence of a virus or worm; repeated unsuccessful login attempts; presence of keylogger or other spyware; and/or malicious rootkits; etc”; ¶29 “the system associates a signal with a risk score indicating a risk or likelihood that one or more events represented by the signal corresponds to a cyber attack.”) based at least in part on an assessment of the AI prompt (Martin, Figure 4); and
in response to determining that the AI prompt is the surreptitious AI prompt (Id), tagging the prompt record to identify the AI prompt as the surreptitious AI prompt as storing the risk score with the signal’s metadata (Martin, ¶30 “the system retrieves a risk score-corresponding to a type of the new signal-from the risk database and stores this risk score with the new signal, such as in the new signal's metadata.”),
[wherein processing the AI prompt to identify whether the AI prompt is the surreptitious prompt comprises] (Please note this claim limitation has been construed as referring to the claimed --assessment--):
generating a prompt vector based at least in part on the AI prompt (Martin, Figure 4, S220 “update/regenerate vector 88677);
comparing the prompt vector to a surreptitious prompt vector generated from the surreptitious prompt to obtain a comparison result (Martin, Figure 4, S270 compare vector to malicious vector database) ; and
identifying whether the AI prompt is the surreptitious prompt based at least in part on the comparison result as generating an alert when match is found (Figure 4, S22).
It would have been obvious to one of ordinary skill to which said subject matter pertains at the time the invention was filed to have performed risk analysis as taught by Martin on the received queries taught by Safronov904 as it yields the predictable results predicting and characterizing if the query contains cyber security threats (Martin, ¶2). One of ordinary skill in the art would recognize that any text input by the user has the possibility of containing security threats, such as malware injections. As such, one of ordinary skill in the art would recognize the ability to evaluate the input data using the methods taught by Martin to determine a risk score for the query being a cyber-attack. Furthermore, one of ordinary skill in the art would further recognize that the system may be customized beyond the specific use cases via the human supervision of the ML classification system (Martin, ¶28).
Claim 43 is rejected under 35 U.S.C. 103 as being unpatentable over Safronov904 [20200192904] in view of Tsubouchi [2017/0270433].
With regard to claim 43 Safronov904 further teaches identifying a request type (Safronov904, ¶93 “A given posting in a given posting list includes some type of data that is indicative of a given document that includes the searchable term associated with the given posting list and, optionally, includes some additional data (for example, where in the document the searchable term appears, number of appearances in the document, and the like).”) of a current generative AI request as the searchable term (Id) as one of a plurality of request types as the type of data (Id); and
[[
Tsubouchi teaches estimating a generation length of the response based on the request type, wherein the selection of the AI prompt among the plurality of AI prompts is further based at least in part on the estimated generation length (Tsubouchi, ¶129 “the generation unit 122 can generate a learning model for each type of a counter terminal behavior, and furthermore, can generate a learning model for each type of context. The generation unit 122 can generate a learning model and can generate a length estimation model for each type of a counter terminal behavior and each type of the two or more contexts”).
It would have been obvious to one of ordinary skill in which said subject matter pertains at the time the invention was filed to have implemented the learning model taught by Safronov904 to use the generation unit for each type of context as taught by Tsubouchi as it yields the predictable results of ensuring that the generated data is context specific (Tsubouchi, ¶130).
Claim 44 is rejected under 35 U.S.C. 103 as being unpatentable over Safronov904 [20200192904] in view of Poulin [2008/0154821].
With regard to claim 44 Safronov904 taches all the limitation of claim 21 as detailed above. Safronov904 does not explicitly teach obtaining user information associated with a user, the user information including project information of a current project associated with the user, wherein the selection of the AI prompt among the plurality of AI prompts is further based at least in part on the user information.
Poulin teaches obtaining user information associated with a user, the user information including project information of a current project associated with the user, wherein the selection of the AI prompt among the plurality of AI prompts is further based at least in part on the user information (Poulin, ¶35 “The model building process 62 is further described in conjunction with the graphical user interfaces mentioned below. The process prompts 72 a user to select a project name that is concatenated with the username and carried throughout the process as a default identifier for the particular project. The process prompts 73 the user to select a categorization for the model to allow easier classification and categorization within the system”).
It would have been obvious to one of ordinary skill in which said subject matter pertains at the time the invention was filed to have implemented the model taught by Safronov904 using the user categorization techniques taught by Poulin as it yields the predictable results of enabling the collected data to be set to specific categories (Poulin, ¶35, ¶41)
Allowable Subject Matter
Claim 42 is objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Ramanujan [2025/0342063] teaches an LLM that determines a model-agnostic per-token latency estimate that enables direct comparison between models relative to latency (See Paragraph [0079]).
Sikand [2026/0064493] teaches latency estimation for an AI prompt, where the runtime latency is based on number of tokens (See paragraph [0055], [0089], [0090]).
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 AMANDA WILLIS whose telephone number is (571)270-7691. The examiner can normally be reached Monday-Friday 8am-2pm.
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, Ajay Bhatia can be reached at 571-272-3906. 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.
/AMANDA L WILLIS/ Primary Examiner, Art Unit 2156
Prosecution Insights