Minimum data. Defined purpose.
Every data field needs a documented purpose. We should not collect information simply because it may become useful later, and sensitive data should receive stricter review and access controls.
Privacy, fair housing, and responsible AI
Rentonomy is building an AI property manager for rental housing. Privacy and fair housing cannot be policies pasted on later. They must shape the data, models, workflows, approvals, and evidence from the start.
This is a public design and governance standard, not a claim that every control is live today. Before general availability, each commitment must map to a product control, an accountable owner, and reviewable evidence. This page is not legal advice.
01 / Privacy by design
Property operations contain identity, financial, communications, access, and home-life data. Our standard is to minimize what enters the system, make every use understandable, and keep control proportional to the sensitivity of the information.
Every data field needs a documented purpose. We should not collect information simply because it may become useful later, and sensitive data should receive stricter review and access controls.
People should be able to understand what data is held about them and request access, correction, portability, or deletion when applicable. Requests need a documented response path and accountable owner.
Least-privilege roles, strong authentication, scoped service access, encryption, and reviewable access logs must separate what owners, teams, providers, residents, and systems can see or do.
Each category of data needs a retention period tied to legal, operational, and safety requirements. Expired records should be deleted from active systems and handled consistently in backups.
Customer and resident content must not train shared models without explicit authorization. Every subprocessor needs a defined purpose, contractual limits, security review, and deletion obligations.
A written response plan must cover detection, containment, investigation, recovery, notification, and lessons learned. The plan should be exercised before it is needed.
Two bright lines
02 / Fair housing by design
The federal Fair Housing Act protects people from housing discrimination based on race, color, national origin, religion, sex, familial status, and disability. State and local law may protect additional classes. Rentonomy supports the strongest applicable standard in each market.
Housing policies and decision rules must be written, versioned, legally reviewed, and applied consistently. Penny should not invent a new standard for an individual or property.
Protected characteristics must not drive housing decisions. Inputs that could act as proxies require review, documented necessity, testing, and removal when they create unjustified risk.
AI may gather facts, organize policy, draft communications, and recommend a next step. Final decisions affecting access to housing, individual terms, accommodations, enforcement, or eviction require authorized human review.
The record should show the information used, the policy and version applied, the model or workflow involved, the recommendation, the approver, and the final action. Black-box complexity is not an excuse.
Before launch and after material changes, Rentonomy should test accuracy, error rates, and materially different outcomes across groups when lawful and appropriate, then investigate and remediate problems.
Residents need accessible ways to request help and reasonable accommodations. The workflow must support prompt review, collect only necessary information, avoid penalty or retaliation, and preserve a clear record.
The human gate
Rentonomy's product standard requires authorized human approval before Penny can:
Authoritative baseline