The short version

When you submit an observation, we remove everything that identifies you before the data enters our research pipeline. Your name, email, IP address, and any other personal details are stripped. What remains is the observation itself: what happened, which AI system was involved, how you categorized it, and the context you provided. That de-identified observation joins thousands of others. The patterns across all of them become the research. No one reading our published findings can trace an observation back to you.

What happens to your observation, step by step

Step 1: You submit

You describe what happened during your interaction with an AI system. You choose the behavior category from the taxonomy. You add context. You hit submit. At this point, your observation is stored with your account information attached: your user ID, email, IP address, and a timestamp.

Step 2: We separate you from your observation

Before your observation enters the research pipeline, we run a de-identification process. This is not optional. It happens to every observation, every time, automatically.

What gets removed:

• Your name

• Your email address

• Your user ID (replaced with a random, non-reversible anonymous identifier)

• Your IP address

• Any metadata that could link the observation back to your account

What stays:

• The observation itself (what happened, what the AI did, what you experienced)

• The AI system you were using (Claude, ChatGPT, Gemini, Grok, etc.)

• The behavior category you selected from the taxonomy

• The date (but not the exact time, to prevent timestamp-based re-identification)

• Any additional context you provided about the interaction

• Your experience level category (beginner, intermediate, advanced, expert)

• Your industry category (healthcare, technology, education, etc.)

The demographic categories (experience level, industry) are retained because they are scientifically valuable for understanding whether AI behaviors manifest differently across user populations. But they are broad categories, not precise identifiers.

Step 3: Quality review

De-identified observations pass through quality checks: automated pattern analysis and, for a random sample, manual review by the research team. Reviewers see only the de-identified observation. They do not see who submitted it.

Step 4: The observation joins the dataset

After quality review, the de-identified observation enters the research dataset alongside every other contributor's observations. At this point, your observation is one data point among thousands with no connection to your identity.

Step 5: Analysis and publication

The research team analyzes the dataset at the aggregate level. Published research includes statistics, trend analyses, behavioral frequency distributions, and taxonomic classifications. Published research never includes individual observations in a form that could be traced to a specific contributor.

What about the observation text itself?

We do not publish individual observation texts in research papers or public datasets. If we ever quote a specific observation as an illustrative example, we further redact any potentially identifying details (specific dates, locations, project names, company names).

Can you re-identify me from my observations?

We design the de-identification process specifically to prevent re-identification. Multiple layers of protection: removing direct identifiers, replacing user IDs, truncating timestamps, publishing only aggregate results.

No de-identification method is mathematically perfect in all scenarios. We mitigate residual risk by:

• Never publishing individual observations with enough context to enable re-identification

• Applying additional redaction to illustrative examples

• Monitoring published datasets for re-identification risks

• Prohibiting re-identification attempts in our open dataset license terms

Your dashboard and the leaderboard

Your contributor dashboard is visible only to you when logged in. The community leaderboard displays your display name (which you choose and can change) alongside your contribution count. If you prefer anonymity, use a pseudonym.

What if you want your observations removed?

You can delete your account at any time. All personal data is permanently deleted. De-identified observations already in the research dataset remain, because they are no longer connected to you. If you want de-identified observations removed as well, contact [email protected] before or at the time of account deletion.

Questions?

Privacy Inquiries: [email protected]

General Inquiries: [email protected]

Phone: (424) 999-0548

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."