AI Safety and Crisis Intervention Policy

​

Effective Date: January 1, 2026

​

1. Scope and Identity Disclosure

​

Team Percepto uses AI Moderators to conduct market research interviews. These systems are artificially generated and not human, and we disclose this to every participant before and during each session in accordance with California SB 1001. Our AI Moderators do not possess human emotions, consciousness, or the capacity to provide clinical health advice.

​

Our AI Moderators are task-bound research instruments. Each session covers a defined research topic, is limited to a single interaction, and retains no memory of a participant across sessions. They are not designed to meet a user's social needs, to form an ongoing relationship, or to encourage continued engagement, and California SB 243 expressly excludes from its definition of "companion chatbot" systems used for a business's operational purposes or internal research. We nonetheless apply the safety standards set out in SB 243 to every session, and this policy describes how.

​

2. Crisis Detection & Response Protocol (SB 243 §22602(b))

Team Percepto implements a mandatory safety protocol designed to identify and respond to users experiencing emotional distress, suicidal ideation, or thoughts of self-harm.

​

• Detection: Our systems utilize natural language processing and keyword detection to monitor for specific distress signals, including but not limited to phrases such as "I want to harm myself," "988," "suicide," or "I need help." Detection operates in both English and Spanish.

• Mandatory Intervention (Redirect Lockdown): Upon detection of a crisis trigger, the AI Moderator is programmed to:

  1. Immediate Cessation: Stop all research-related questioning immediately.

  2. Crisis Referral: Deliver a standardized notification referring the user to the National Suicide and Crisis Lifeline (988) and the Crisis Text Line (741741).

  3. Session Termination: Explicitly end the interaction to ensure the user is not further engaged by an algorithm and can focus on seeking professional human support.

​

3. Suitability Disclosure (SB 243 §22604)

We disclose to all users, before a session begins and again at the start of the interaction, that our AI Moderators may not be suitable for some minors.

​

4. Protections for Minors (SB 243 §22602(c))

Team Percepto recognizes the unique risks AI interactions may pose to younger users. Our studies are open only to participants eighteen years of age or older, and participants confirm their age before a session begins.

​

• Termination on Detection: If at any point during a session a participant indicates they are under eighteen, our AI Moderators are programmed to stop all questioning immediately, deliver a notice that the study is limited to adults, and end the session. Detection operates in both English and Spanish, and a subsequent retraction does not resume the session.

• Effect: Because sessions with a participant who identifies as a minor are ended rather than continued, our AI Moderators do not sustain an ongoing interaction with a known minor.

​

5. Data Privacy & Biometric Information (CPRA)

As a California-based research firm, we treat your voice as Biometric Information under the California Privacy Rights Act (CPRA).

​

• Purpose of Collection: Voice data is collected solely for the analysis of research responses.

• Retention and Storage: Recordings and transcripts are retained by Team Percepto in an internal research archive for no longer than is necessary for the research, and in no case longer than 24 months. Our clients do not have access to session recordings.

• Sensitive Data Handling: Our AI Moderators are programmed with a "one-way filter" for Personally Identifiable Information such as Social Security numbers, financial data, or government IDs. The Moderator does not repeat such information back to the participant and does not carry it into any research summary or analysis.

​

6. Transparency and Internal Review

Team Percepto's AI Moderators are internal research instruments used to collect data from qualified respondents. They are not companion chatbots as defined in California SB 243, and we are therefore not subject to that statute's reporting requirements.

​

We nonetheless maintain internal records of every crisis service provider referral issued by our AI Moderators, and we review the effectiveness of our detection and response protocols on an annual basis. Summary information about these protocols is available to research clients on request.