Data Collection
“Trust starts with restraint. We built AETHEON so the substance of what you ask, and what our models say back, is never stored, never logged, and never used to train anyone's model. This policy is how we keep that promise visible and accountable.”
Marie Gill
Founder & CEO, AETHEON
A privacy-by-design commitment
This Data Collection Policy describes the categories of data that AETHEON, Inc. ("AETHEON," "we") collects, and does not collect, when individuals and organizations use the large language model ("LLM") and related artificial intelligence features within our products and services (collectively, the "Services").
AETHEON is built on a privacy-by-design principle: we do not store or log the content of transient LLM prompts or model responses. We collect only operational metadata necessary to run, secure, bill for, and improve the performance of those interactions.
This Policy governs the transient LLM interaction layer across the Services. It does not govern records a user intentionally submits for storage as part of a requested product workflow, including profile information, job applications, résumés, published content, or account records. Those records may be retained to provide the requested feature and are subject to the applicable product notices, controls, and retention terms.
Terms used in this policy
| Term | Meaning |
|---|---|
| Prompt | Any input submitted to the Services by or on behalf of a user, including text, instructions, attachments, and embedded context. |
| Response | Any output generated by an LLM or the Services in reply to a prompt. |
| Content Data | The substance of transient prompts and responses exchanged with an LLM. It does not include profile, application, résumé, or other records a user intentionally submits for storage as part of a requested product workflow. |
| Metadata | Operational and performance data about an interaction that does not include the substance of the prompt or response. |
| Retention | The persistence of data beyond the moment it is needed to fulfill a request, in any durable store, log, cache, or training corpus. |
No content, anywhere
AETHEON does not store, log, or otherwise retain Content Data from transient LLM interactions. Specifically:
- We do not store the content of transient LLM prompts or responses in any database, file system, object store, or other durable medium.
- We do not log the content of transient LLM prompts or responses in application logs, debug logs, audit logs, error traces, or any other logging system.
- We use provider data controls and contractual terms designed to prevent prompts and responses from being used for model training, fine-tuning, or unrelated human review.
- We do not build user profiles, advertising segments, or behavioral models from the substance of prompts or responses.
Content Data exists only transiently in memory for the time strictly necessary to process a request and return a response, after which it is discarded and not persisted.
Operational telemetry, not substance
To operate, secure, bill for, and improve the performance of the Services, we collect operational metadata that describes an interaction without capturing its substance.
| Metadata category | Description |
|---|---|
| Prompt token count | The number of tokens contained in an input prompt. |
| Completion token count | The number of tokens contained in a generated response. |
| Response time | The total time taken to generate and return a response. |
| Latency | Time-to-first-token and other latency measurements across the request lifecycle. |
| Cost | The computed cost attributable to an interaction, derived from token counts and model pricing. |
| Other performance metrics | Model identifier, request status and error codes, throughput, rate-limit usage, and similar operational telemetry. |
This metadata is computed from, but does not contain, Content Data. For example, a token count records how many tokens were processed without recording what those tokens were.
What we use metadata for
- Operating and maintaining the reliability and availability of the Services.
- Monitoring and improving performance, latency, and throughput.
- Calculating usage and billing, and providing usage reporting to customers.
- Capacity planning, cost management, and resource allocation.
- Detecting, investigating, and preventing abuse, fraud, and security incidents.
We do not use metadata to reconstruct, infer, or approximate the substance of prompts or responses.
Zero-retention by contract and configuration
Where the Services rely on third-party LLM providers, AETHEON uses available provider data controls and contractual terms designed to minimize retention and prevent prompts and responses from being used to train or improve general-purpose models. Provider capabilities and legally required security retention can vary. We periodically review provider configurations, terms, and attestations to confirm that each integration continues to meet our privacy requirements.
How we hold the data we do keep
7.1 Retention
Content Data from transient LLM interactions is not retained by AETHEON. Intentionally saved product records are outside this Policy and may be retained to provide the requested workflow. Metadata is retained only for as long as necessary to fulfill the purposes above, after which it is deleted or aggregated into non-identifying statistics in accordance with our internal retention schedule.
7.2 Security
AETHEON protects collected metadata using administrative, technical, and physical safeguards appropriate to its sensitivity, including access controls, encryption in transit and at rest, least-privilege access, and audit logging of administrative access to metadata systems.
Owned, reviewed, enforced
This Policy is owned by AETHEON's security and privacy function and is reviewed at least annually, and whenever there is a material change to the Services or applicable law. Engineering changes that could affect the collection, logging, or retention of Content Data must undergo privacy review before release. Violations of this Policy are treated as security incidents and handled under our incident response process.
Questions about this Policy may be directed to AETHEON's privacy team at privacy@aetheon.ai, or sent through our contact form.