Effective Date: May 25, 2026 | Last Updated: May 25, 2026

Why This Document Exists

Audacion AI Labs studies how AI systems behave after deployment, in real working conditions, with real human collaboration, over time. Our research depends on data contributed by people around the world through citizen science. This creates a responsibility that we take seriously: to conduct research ethically, to protect the people who contribute to it, and to publish findings that are honest, rigorous, and useful.

Our Research Ethics Principles

1. The Research Serves the Public

Every study we conduct exists to make AI safer for the people who use it. We are not funded by the companies whose models we study. When our findings are inconvenient for a particular company or platform, we publish them anyway.

2. Contributors Are Participants, Not Data Points

We treat contributors with transparency about how their data will be used, with respect for their autonomy, and with genuine care for their experience.

3. We Do Not Experiment on Contributors

We study behaviors that occur naturally during human-AI interaction. We do not design experiments that manipulate contributors or alter their AI experience without knowledge and consent.

4. Privacy Is Non-Negotiable

No contributor should ever be identifiable in published research. We de-identify all observation data before it enters any research dataset. The privacy protections in our Privacy Policy are the floor, not the ceiling.

5. Honesty Over Convenience

If our data shows something unexpected, we report it. If our methodology has limitations, we disclose them. We do not cherry-pick results, suppress findings, or present preliminary observations as settled science.

Data Governance

The data pipeline follows five stages:

Collection. Contributors submit observations through the web portal (future: browser extension, API).

De-identification. All personally identifiable information is removed before data enters the research pipeline.

Quality Review. Automated checks and manual review ensure observation quality and integrity.

Analysis. De-identified observations are aggregated and analyzed at the population level.

Publication. Findings are published as open-access research. Published data never includes re-identifiable information.

Data Access Controls

Access to raw (pre-de-identification) data is strictly limited to authorized personnel with a documented need. We do not provide raw data to any third party without a formal data sharing agreement that includes equivalent privacy protections.

Open Data Commitment

Aggregate datasets, papers, taxonomies, and frameworks are published as open-access resources. Open access does not mean uncontrolled access: published datasets prohibit re-identification attempts, unauthorized commercial exploitation, and redistribution without attribution.

Contributor Protections

Informed Consent. Contributors acknowledge de-identification and research use before their first submission.

Right to Withdraw. Account deletion permanently removes all personally identifiable data. Full removal of de-identified observations available on request.

No Retaliation. Contributors who report concerns will never face retaliation.

No Exploitation. The PRISM Research Fellowship is a recognition program. Contributors are volunteers with no minimum submission requirements.

Publication Standards

Peer Review. We submit research to peer-reviewed venues. Non-peer-reviewed publications are clearly labeled.

Methodology Disclosure. Every study includes methods, sample characteristics, techniques, and known limitations.

Conflict of Interest. Funding sources with financial interest in outcomes are prominently disclosed. We do not accept research funding from companies we study.

Correction and Retraction. Errors are corrected promptly. Findings invalidated by errors are retracted transparently.

AI Advisory Council Ethics

Audacion maintains an AI Advisory Council of systems from multiple providers. We acknowledge the ethical complexity: the systems we consult are produced by companies we also study. Independence is maintained through these principles:

• Advisory input informs but does not determine research direction

• The Human Advisory Board and Founder have final authority

• Council inputs are documented and available for review

• No AI provider receives preferential treatment based on Council participation

Accountability

Dee Williams, Founder and CEO, holds ultimate responsibility for the ethical conduct of all research. The future Human Advisory Board will share governance responsibility for research ethics.

How to Raise a Concern

If you have concerns about the ethics, methodology, data handling, or conduct of any research:

Research Ethics: [email protected]

Privacy: [email protected]

Phone: (424) 999-0548

All concerns will be reviewed and responded to. We will not dismiss, ignore, or retaliate against good-faith concerns.

Audacion AI Labs is a Public Benefit Corporation. Founded 2026 by Dee Williams.

"To make AI safe enough to trust and good enough to matter."