Systems and methods for dynamic de-identification and anonymity for user input to artificial intelligence (AI)-based models
Various systems, computer-readable media, and computer-implemented methods of providing improved data privacy, anonymity and security by enabling subjects to which data pertains to remain “dynamically anonymous,” i.e., anonymous for as long as is desired—and to the extent that is desired—are disclosed herein. Embodiments include systems that create, access, use, store and/or erase data with increased privacy, anonymity, and security—thereby facilitating the availability of more qualified and accurate information. When personal data is authorized by data subjects to be shared with third parties, embodiments described herein may facilitate the sharing of information in a dynamically-controlled manner that also enables the delivery of temporally-, geographically-, and/or purpose-limited information to the receiving party. In one example, the disclosed techniques may be used to functionally separate geospatial information, such that it remains “dynamically anonymous,” i.e., anonymous for as long as is desired—and to the extent or degree that is desired.
1. A system, comprising:
a communication interface for sending data;
one or more memory elements having, stored therein, computer program code; and
one or more processing units operatively coupled to the one or more memory elements and configured to execute instructions in the computer program code that cause the one or more processing units to:
receive, from a first client, a first input to an artificial intelligence (AI) model;
identify, within the first input, a first collection of one or more sensitive data elements;
generate, for each sensitive data element in the first collection of sensitive data elements, a different dynamically-changing, temporally unique identifier;
associate each generated dynamically-changing, temporally unique identifier with the sensitive data element from the first collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier was generated for;
generate a first de-identified version of the first input to the AI model by replacing, in the first input, each occurrence of a sensitive data element from the first collection of sensitive data elements with the respective associated dynamically-changing, temporally unique identifier; and
store, in at least one of the one or more memory elements, for the first collection of sensitive data elements, information that, when used in combination with an associated dynamically-changing, temporally unique identifier, is needed to recover the sensitive data element from the first collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier is associated with.
2. The system of claim 1 , wherein the one or more processing units are further configured to execute instructions in the computer program code that cause the one or more processing units to:
receive a second input to the AI model;
identify, within the second input, a second collection of one or more sensitive data elements, wherein the second collection of sensitive data elements has at least one data element in common with the first collection of sensitive data elements;
generate, for each sensitive data element in the second collection of sensitive data elements, a different dynamically-changing, temporally unique identifier, wherein at least one of the dynamically-changing, temporally unique identifiers associated with the second collection of sensitive data elements is the same as one of the dynamically-changing, temporally unique identifiers associated with the first collection of sensitive data;
associate each generated different dynamically-changing, temporally unique identifier with the sensitive data element from the second collection of sensitive data elements that the respective different dynamically-changing, temporally unique identifier was generated for;
generate a second de-identified version of the second input to the AI model by replacing, in the second input, each occurrence of a sensitive data element from the second collection of sensitive data elements with the respective associated dynamically-changing, temporally unique identifier; and
store, in at least one of the one or more memory elements, for the second collection of sensitive data elements, information that, when used in combination with an associated dynamically-changing, temporally unique identifier, is needed to recover the sensitive data element from the second collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier is associated with.
3. The system of claim 1 , wherein the AI model comprises a generative AI model.
4. The system of claim 2 , wherein the second input is received from one of: the first client; or a second client.
5. The system of claim 1 , wherein the one or more processing units are further configured to execute instructions in the computer program code that cause the one or more processing units to:
send, to the first client, the first de-identified version of the first input to the AI model.
6. The system of claim 4 , wherein the one or more processing units are further configured to execute instructions in the computer program code that cause the one or more processing units to:
(a) when the second input is received from the first client, send, to the first client, the second de-identified version of the second input to the AI model; or
(b) when the second input is received from the second client, send, to the second client, the second de-identified version of the second input to the AI model.
7. The system of claim 2 , wherein the one or more processing units are further configured to execute instructions in the computer program code that cause the one or more processing units to:
receive, from the first client, a first de-identified response from the AI model resulting from processing the first de-identified version of the first input;
identify, within the first de-identified response, a third collection of dynamically-changing, temporally unique identifiers;
retrieve, from the at least one of the one or more memory elements, the information that when used in combination with the third collection of dynamically-changing, temporally unique identifiers, is needed to recover the associated sensitive data elements from the first collection of sensitive data elements;
recover, for each dynamically-changing, temporally unique identifier in the third collection of dynamically-changing, temporally unique identifiers, the respective sensitive data element it is associated with in the first collection of sensitive data elements;
associate each sensitive data element from the first collection of sensitive data elements with the respective dynamically-changing, temporally unique identifier in the third collection of dynamically-changing, temporally unique identifiers; and
generate a first re-identified response by replacing, in the first de-identified response, each occurrence of a dynamically-changing, temporally unique identifier from the third collection of dynamically-changing, temporally unique identifiers with the respective associated sensitive data element.
8. The system of claim 7 , wherein the one or more processing units are further configured to execute instructions in the computer program code that cause the one or more processing units to:
send the first re-identified response to at least one of: the first client; a second client; or another client.
9. The system of claim 7 , wherein the one or more processing units are further configured to execute instructions in the computer program code that cause the one or more processing units to:
receive, from the first client or a second client, a second de-identified response from the AI model resulting from processing the second de-identified version of the second input;
identify, within the second de-identified response, a fourth collection of dynamically-changing, temporally unique identifiers;
retrieve, from the at least one of the one or more memory elements, the information that when used in combination with the fourth collection of dynamically-changing, temporally unique identifiers, is needed to recover the associated sensitive data elements from the second collection of sensitive data elements;
recover, for each dynamically-changing, temporally unique identifier in the fourth collection of dynamically-changing, temporally unique identifiers, the respective sensitive data element it is associated with in the second collection of sensitive data elements;
associate each sensitive data element from the second collection of sensitive data elements with the respective dynamically-changing, temporally unique identifier in the fourth collection of dynamically-changing, temporally unique identifiers; and
generate a second re-identified response by replacing, in the second de-identified response, each occurrence of a dynamically-changing, temporally unique identifier from the fourth collection of dynamically-changing, temporally unique identifiers with the respective associated sensitive data element.
10. The system of claim 9 , wherein the one or more processing units are further configured to execute instructions in the computer program code that cause the one or more processing units to:
send the second re-identified response to at least one of: the first client; the second client; or another client.
11. A non-transitory computer readable medium comprising computer executable instructions stored thereon to cause one or more processing units to:
receive, from a first client, a first input to an artificial intelligence (AI) model;
identify, within the first input, a first collection of one or more sensitive data elements;
generate, for each sensitive data element in the first collection of sensitive data elements, a different dynamically-changing, temporally unique identifier;
associate each generated dynamically-changing, temporally unique identifier with the sensitive data element from the first collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier was generated for;
generate a first de-identified version of the first input to the AI model by replacing, in the first input, each occurrence of a sensitive data element from the first collection of sensitive data elements with the respective associated dynamically-changing, temporally unique identifier; and
store, in at least one memory element, for the first collection of sensitive data elements, information that, when used in combination with an associated dynamically-changing, temporally unique identifier, is needed to recover the sensitive data element from the first collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier is associated with.
12. The non-transitory computer readable medium of claim 11 , wherein the instructions stored thereon to further cause the one or more processing units to:
receive a second input to the AI model;
identify, within the second input, a second collection of one or more sensitive data elements, wherein the second collection of sensitive data elements has at least one data element in common with the first collection of sensitive data elements;
generate, for each sensitive data element in the second collection of sensitive data elements, a different dynamically-changing, temporally unique identifier, wherein at least one of the dynamically-changing, temporally unique identifiers associated with the second collection of sensitive data elements is the same as one of the dynamically-changing, temporally unique identifiers associated with the first collection of sensitive data;
associate each generated different dynamically-changing, temporally unique identifier with the sensitive data element from the second collection of sensitive data elements that the respective different dynamically-changing, temporally unique identifier was generated for;
generate a second de-identified version of the second input to the AI model by replacing, in the second input, each occurrence of a sensitive data element from the second collection of sensitive data elements with the respective associated dynamically-changing, temporally unique identifier; and
store, in at least one memory element, for the second collection of sensitive data elements, information that, when used in combination with an associated dynamically-changing, temporally unique identifier, is needed to recover the sensitive data element from the second collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier is associated with.
13. The non-transitory computer readable medium of claim 12 , wherein the first collection of one or more sensitive data elements and the second collection of one or more sensitive data elements each comprises:
one or more words, numbers, phrases, characters, or symbols representing personal, identifying, or proprietary information.
14. The non-transitory computer readable medium of claim 12 , wherein the second input is received from one of: the first client; or a second client.
15. The non-transitory computer readable medium of claim 11 , wherein the instructions stored thereon to further cause the one or more processing units to:
send, to the first client, the first de-identified version of the first input to the AI model.
16. The non-transitory computer readable medium of claim 14 , wherein the instructions stored thereon to further cause the one or more processing units to:
(a) when the second input is received from the first client, send, to the first client, the second de-identified version of the second input to the AI model to the first client when the second input is received from the first client; or
(b) when the second input is received from the second client, send, to the second client, the second de-identified version of the second input to the AI model.
17. The non-transitory computer readable medium of claim 12 , wherein the instructions stored thereon to further cause the one or more processing units to:
receive, from the first client, a first de-identified response from the AI model resulting from processing the first de-identified version of the first input;
identify, within the first de-identified response, a third collection of dynamically-changing, temporally unique identifiers;
retrieve, from the at least one of the one or more memory elements, the information that when used in combination with the third collection of dynamically-changing, temporally unique identifiers, is needed to recover the associated sensitive data elements from the first collection of sensitive data elements;
recover, for each dynamically-changing, temporally unique identifier in the third collection of dynamically-changing, temporally unique identifiers, the respective sensitive data element it is associated with in the first collection of sensitive data elements;
associate each sensitive data element from the first collection of sensitive data elements with the respective dynamically-changing, temporally unique identifier in the third collection of dynamically-changing, temporally unique identifiers; and
generate a first re-identified response by replacing, in the first de-identified response, each occurrence of a dynamically-changing, temporally unique identifier from the third collection of dynamically-changing, temporally unique identifier with the respective associated sensitive data element.
18. The non-transitory computer readable medium of claim 17 , wherein the instructions stored thereon to further cause the one or more processing units to:
send the first re-identified response to at least one of: the first client; a second client; or another client.
19. The non-transitory computer readable medium of claim 17 , wherein the instructions stored thereon to further cause the one or more processing units to:
receive, from the first client or a second client, a second de-identified response from the AI model resulting from processing the second de-identified version of the second input;
identify, within the second de-identified response, a fourth collection of dynamically-changing, temporally unique identifiers;
retrieve, from the at least one of the one or more memory elements, the information that when used in combination with the fourth collection of dynamically-changing, temporally unique identifiers, is needed to recover the associated sensitive data elements from the second collection of sensitive data elements;
recover, for each dynamically-changing, temporally unique identifier in the fourth collection of dynamically-changing, temporally unique identifiers, the respective sensitive data element it is associated with in the second collection of sensitive data elements;
associate each sensitive data element from the second collection of sensitive data elements with the respective dynamically-changing, temporally unique identifier in the fourth collection of dynamically-changing, temporally unique identifiers; and
generate a second re-identified response by replacing, in the second de-identified response, each occurrence of a dynamically-changing, temporally unique identifier from the fourth collection of dynamically-changing, temporally unique identifier with the respective associated sensitive data element.
20. The non-transitory computer readable medium of claim 19 , wherein the instructions stored thereon to further cause the one or more processing units to:
send the second re-identified response to at least one of: the first client; the second client; or another client.
21. A method, comprising:
receiving, from a first client, a first input to an artificial intelligence (AI) model;
identifying, within the first input, a first collection of one or more sensitive data elements;
generating, each sensitive data element in the first collection of sensitive data elements, a different dynamically-changing, temporally unique identifier;
associating each generated dynamically-changing, temporally unique identifier with the sensitive data element from the first collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier was generated for;
generating a first de-identified version of the first input to the AI model by replacing, in the first input, each occurrence of a sensitive data element from the first collection of sensitive data elements with the respective associated dynamically-changing, temporally unique identifier; and
storing, in at least one memory element, for the first collection of sensitive data elements, information that, when used in combination with an associated dynamically-changing, temporally unique identifier, is needed to recover the sensitive data element from the first collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier is associated with.
22. The method of claim 21 , further comprising:
receiving a second input to the AI model;
identifying, within the second input, a second collection of one or more sensitive data elements, wherein the second collection of sensitive data elements has at least one data element in common with the first collection of sensitive data elements;
generating, for each sensitive data element in the second collection of sensitive data elements, a different dynamically-changing, temporally unique identifier, wherein at least one of the dynamically-changing, temporally unique identifiers associated with the second collection of sensitive data elements is the same as one of the dynamically-changing, temporally unique identifiers associated with the first collection of sensitive data;
associating each generated different dynamically-changing, temporally unique identifier with the sensitive data element from the second collection of sensitive data elements that the respective different dynamically-changing, temporally unique identifier was generated for;
generating a second de-identified version of the second input to the AI model by replacing, in the second input, each occurrence of a sensitive data element from the second collection of sensitive data elements with the respective associated dynamically-changing, temporally unique identifier; and
storing, in at least one memory element, for the second collection of sensitive data elements, information that, when used in combination with an associated dynamically-changing, temporally unique identifier, is needed to recover the sensitive data element from the second collection of sensitive data elements that the respective dynamically-changing, temporally unique identifier is associated with.
23. The method of claim 22 , wherein the first collection of one or more sensitive data elements and the second collection of one or more sensitive data elements each comprises:
one or more words, numbers, phrases, characters, or symbols representing personal, identifying, or proprietary information.
24. The method of claim 22 , wherein the second input is received from one of: the first client; or a second client.
25. The method of claim 21 , further comprising:
sending, to the first client, the first de-identified version of the first input to the AI model.
26. The method of claim 24 , further comprising:
(a) when the second input is received from the first client, sending, to the first client, the second de-identified version of the second input to the AI model; or
(b) when the second input is received from the second client, sending, to the second client, the second de-identified version of the second input to the AI model.
27. The method of claim 22 , further comprising:
receiving, from the first client, a first de-identified response from the AI model resulting from processing the first de-identified version of the first input;
identifying, within the first de-identified response, a third collection of dynamically-changing, temporally unique identifiers;
retrieving, from the at least one of the one or more memory elements, the information that when used in combination with the third collection of dynamically-changing, temporally unique identifiers, is needed to recover the associated sensitive data elements from the first collection of sensitive data elements;
recovering, for each dynamically-changing, temporally unique identifier in the third collection of dynamically-changing, temporally unique identifiers, the respective sensitive data element it is associated with in the first collection of sensitive data elements;
associating each sensitive data element from the first collection of sensitive data elements with the respective dynamically-changing, temporally unique identifier in the third collection of dynamically-changing, temporally unique identifiers; and
generating a first re-identified response by replacing, in the first de-identified response, each occurrence of a dynamically-changing, temporally unique identifier from the third collection of dynamically-changing, temporally unique identifier with the respective associated sensitive data element.
28. The method of claim 27 , further comprising:
sending the first re-identified response to at least one of: the first client; a second client; or another client.
29. The method of claim 27 , further comprising:
receiving, from the first client or a second client, a second de-identified response from the AI model resulting from processing the second de-identified version of the second input;
identifying, within the second de-identified response, a fourth collection of dynamically-changing, temporally unique identifiers;
retrieving, from the at least one of the one or more memory elements, the information that when used in combination with the fourth collection of dynamically-changing, temporally unique identifiers, is needed to recover the associated sensitive data elements from the second collection of sensitive data elements;
recovering, for each dynamically-changing, temporally unique identifier in the fourth collection of dynamically-changing, temporally unique identifiers, the respective sensitive data element it is associated with in the second collection of sensitive data elements;
associating each sensitive data element from the second collection of sensitive data elements with the respective dynamically-changing, temporally unique identifier in the fourth collection of dynamically-changing, temporally unique identifiers; and
generating a second re-identified response by replacing, in the second de-identified response, each occurrence of a dynamically-changing, temporally unique identifier from the fourth collection of dynamically-changing, temporally unique identifier with the respective associated sensitive data element.
30. The method of claim 29 , further comprising:
sending the second re-identified response to at least one of: the first client; the second client; or another client.