# US 2025/0378019 A1 · Systems and methods for scaling artificial intelligence memories

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2025/0378019 A1
- Document ID: US20250378019A1
- Status: Published application
- Kind: A1
- Application number: 19/291,271
- Filed: 2025-08-05
- Published: 2025-12-11
- Earliest claimed: 2024-06-06
- Assignees: Blumira Inc.
- Applicants: Blumira Inc.
- Inventors: Matthew Warner
- Web page: https://patentgrep.com/patent/US20250378019A1

## CPC

- G06F 12/023: {Free address space management}
- G06F 2212/1044: Space efficiency improvement
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

The present disclosure pertains to systems and methods for scaling artificial intelligence (AI) memories, addressing storage and relevancy in generative AI frameworks. The described aspects involve an approach for memory management where event summaries and contextual metadata are stored and memories are compressed to conserve storage space while retaining significant information. Various other methods and systems are also disclosed.

## Claims

### Claim 1

A computer-implemented method comprising:
storing, within a storage subsystem of a generative artificial intelligence system, structured memory representation comprising vectorized and semantic event data, wherein the memory of the event comprises a summary of the event and context associated with the event;
determining that the memory has decreased in importance for use in responding to prompts provided to the generative artificial intelligence system;
in response to the determination that the memory has decreased in importance, compressing the memory such that the compressed memory uses less storage space in the storage subsystem than an uncompressed memory of the event;
replacing the uncompressed memory with the compressed memory in the storage subsystem; and
using the compressed memory to respond to a prompt provided to the generative artificial intelligence system.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
storing, within a storage subsystem of a generative artificial intelligence system, a memory of an additional event, wherein:
the memory of the additional event comprises context indicating that the memory is a core memory; and
the memory of the event comprises a non-core memory; and
the memory of the additional event has greater importance than the memory of the event and takes up more storage space in the storage subsystem than the memory of the event.

### Claim 3 (depends on 1)

The method of claim 1, wherein:
the event comprises a security event; and
storing the memory of the event comprises evaluating the event by a plurality of artificial intelligence security operations center agents to identify the context and create the summary.

### Claim 4 (depends on 1)

The method of claim 1, wherein compressing the compressed memory comprises quantizing a vector that stores the summary of the event.

### Claim 5 (depends on 1)

The method of claim 1, wherein determining that the memory has decreased in importance comprises evaluating at least one of: a time since the memory was last accessed, a time since the memory was created, or a relevancy tag associated with the memory.

### Claim 6 (depends on 1)

The method of claim 1, wherein the memory is a shared memory accessible by a plurality of artificial intelligence agents, each agent having independent contextual memory.

### Claim 7 (depends on 1)

The method of claim 1, wherein using the compressed memory to respond to a prompt comprises injecting the compressed memory into a context window of a generative artificial intelligence model.

### Claim 8 (depends on 1)

The method of claim 1, wherein compressing the memory comprises applying a dynamic compression factor based on at least one of: available storage space, relevancy of the memory, or a predefined compression floor.

### Claim 9 (depends on 1)

The method of claim 1, wherein replacing the uncompressed memory with the compressed memory comprises maintaining a reference to an original context of the memory to enable navigation and searching of the memory.

### Claim 10

A system comprising:
one or more physical processors;
physical memory comprising computer-executable instructions that, when executed by the one or more physical processors, cause the one or more physical processors to:
store, within a storage subsystem of a generative artificial intelligence system, memory of an event, wherein the memory of the event comprises a summary of the event and context associated with the event, and
determine that the memory has decreased in importance for use in responding to prompts provided to a generative artificial intelligence system,
in response to the determination that the memory has decreased in importance, compress the memory such that the compressed memory uses less storage space in the storage subsystem than an uncompressed memory of the event, and
replace the uncompressed memory with the compressed memory in the storage subsystem; and
use the compressed memory to respond to a prompt provided to the generative artificial intelligence system.

### Claim 11 (depends on 10)

The system of claim 10, wherein the computer-executable instructions, when executed by at least one of the one or more physical processors, further cause the one or more physical processors to:
store, within a storage subsystem of a generative artificial intelligence system, a memory of an additional event, wherein:
the memory of the additional event comprises context indicating that the memory is a core memory; and
the memory of the event comprises a non-core memory; and
the memory of the additional event has greater importance than the memory of the event and takes up more storage space in the storage subsystem than the memory of the event.

### Claim 12 (depends on 10)

The system of claim 10, wherein:
the event comprises a security event; and
the computer-executable instructions cause the one or more physical processors to store the memory of the event by evaluating the event by a plurality of artificial intelligence security operations center agents to identify the context and create the summary.

### Claim 13 (depends on 10)

The system of claim 10, wherein the computer-executable instructions cause the one or more physical processors to compress the compressed memory by quantizing a vector that stores the summary of the event.

### Claim 14 (depends on 10)

The system of claim 10, wherein the computer-executable instructions cause the one or more physical processors to determine that the memory has decreased in importance by evaluating at least one of: a time since the memory was last accessed, a time since the memory was created, or a relevancy tag associated with the memory.

### Claim 15 (depends on 10)

The system of claim 10, wherein the memory is a shared memory accessible by a plurality of artificial intelligence agents, each agent having independent contextual memory.

### Claim 16 (depends on 10)

The system of claim 10, wherein the computer-executable instructions cause the one or more physical processors to use the compressed memory to respond to a prompt by injecting the compressed memory into a context window of a generative artificial intelligence model.

### Claim 17 (depends on 10)

The system of claim 10, wherein the computer-executable instructions cause the one or more physical processors to compress the memory by applying a dynamic compression factor based on at least one of: available storage space, relevancy of the memory, or a predefined compression floor.

### Claim 18 (depends on 10)

The system of claim 10, wherein the computer-executable instructions cause the one or more physical processors to replace the uncompressed memory with the compressed memory by maintaining a reference to an original context of the memory to enable navigation and searching of the memory.

### Claim 19

A non-transitory computer-readable medium comprising computer-executable instructions that, when executed by one or more physical processors of a computing device, cause the computing device to:
store, within a storage subsystem of a generative artificial intelligence system, memory of an event, wherein the memory of the event comprises a summary of the event and context associated with the event, and
determine that the memory has decreased in importance for use in responding to prompts provided to a generative artificial intelligence system,
in response to the determination that the memory has decreased in importance, compress the memory such that the compressed memory uses less storage space in the storage subsystem than an uncompressed memory of the event, and
replace the uncompressed memory with the compressed memory in the storage subsystem; and
use the compressed memory to respond to a prompt provided to the generative artificial intelligence system.

### Claim 20 (depends on 19)

The non-transitory computer-readable medium of claim 19, wherein the computer-executable instructions, when executed by the one or more physical processors, further cause the one or more physical processors to:
store, within a storage subsystem of a generative artificial intelligence system, a memory of an additional event, wherein:
the memory of the additional event comprises context indicating that the memory is a core memory; and
the memory of the event comprises a non-core memory; and
the memory of the additional event has greater importance than the memory of the event and takes up more storage space in the storage subsystem than the memory of the event.
