Second Innings AI Features Disclosure and Crisis Safety Protocol
Second Innings, Inc.
Last updated November 1, 2026
IMPORTANT NOTICES - PLEASE READ
- THE AI COMPANION IS AN AUTOMATED ARTIFICIAL INTELLIGENCE SYSTEM. IT IS NOT A HUMAN, NOT SECOND INNINGS STAFF, AND NOT ANOTHER MEMBER.
- AI FEATURES ARE NOT A THERAPIST, COUNSELOR, DOCTOR, OR CRISIS SERVICE AND DO NOT PROVIDE MEDICAL, MENTAL HEALTH, LEGAL, FINANCIAL, OR OTHER PROFESSIONAL ADVICE.
- IF YOU ARE THINKING ABOUT SUICIDE OR SELF-HARM, OR ARE IN DANGER: IN THE US, CALL OR TEXT 988 (SUICIDE AND CRISIS LIFELINE) OR TEXT HOME TO 741741 (CRISIS TEXT LINE). IN AN EMERGENCY, CALL 911 (US AND CANADA), 112 (EU, UK AND INDIA), 999 (UK), 000 (AUSTRALIA), OR YOUR LOCAL EMERGENCY NUMBER.
- AI OUTPUTS MAY BE WRONG. Check anything important before relying on it.
- COMPANION CHATBOTS MAY NOT BE SUITABLE FOR SOME MINORS. Second Innings is only for adults aged 18 and over.
1. ABOUT THIS DISCLOSURE
1.1 This AI Features Disclosure and Crisis Safety Protocol (this "Disclosure") explains which features of the Services use artificial intelligence and other automated technologies, how they work, what they can and cannot do, how we handle the data they use, and the protocol we follow when a member expresses thoughts of suicide or self-harm to the AI Companion. It includes the crisis safety protocol that California Senate Bill 243 (2025) requires operators of companion chatbots to publish on their website, and our transparency statement under Article 50 of the EU Artificial Intelligence Act.
1.2 This Disclosure is incorporated into our Terms of Service and forms part of the "Agreement" described in Section 1.3 of the Terms. It supplements Section 9 (AI Features) and Section 10 (Astrology, Compatibility, and Match Insights) of the Terms and does not limit them. Capitalized terms not defined here, such as "AI Features," "Inputs," "Insight Features," and "Informational Content," have the meanings given in the Terms. Our Privacy Policy and Biometric Data Policy explain how we handle personal information and Biometric Data.
2. HOW WE USE AI AND AUTOMATION
2.1 A mix of technologies. As described in Section 9.1 of the Terms, AI Features may use machine learning, large language models, computer vision, and other automated or algorithmic technologies, alone or together with rule-based systems, human-designed logic, scoring formulas, and pre-written templates. Many features combine these. For example, a feature may use rules to select and rank candidates, a learned model to estimate preferences, and templates to explain the result in words.
2.2 What "AI" means on Second Innings. We use labels such as "AI," "AI-powered," "AI insight," or "AI Matchmaker" for features that are built on these automated technologies, including features where some or all of the output is produced by rule-based or template-based components working together with learned models (Terms Section 9.1). Section 3 describes what each feature does so that you can decide how much weight to give it.
2.3 Providers. Some AI Features use models provided by third parties, and some use models that we run ourselves. Our current large language model providers are OpenRouter, which routes our requests to models from OpenAI, and, where configured, OpenAI directly. Models, providers, and techniques may change without notice (Terms Section 9.8).
3. OUR AI FEATURES
The AI Features currently offered may include the following. Availability may depend on your plan, add-ons, location, platform, and app version, and may be subject to usage limits.
| Feature | What it does | How it works |
|---|---|---|
| AI Companion (also called AI Buddy or AI Assistant) | A chat assistant for icebreakers, profile tips, dating questions, and general conversation | Your message is sent to a large language model through our servers, with instructions we write. Replies are generated automatically. It does not remember past conversations. Crisis language triggers the protocol in Section 9. Daily message limits apply. |
| Matching and recommendations (including Smart Match, Nearby, and the card stack) | Chooses which profiles you see and in what order | Rule-based filters and scoring using your preferences, filters, profile, questionnaire answers, location, and activity, together with preference signals learned from the profiles you like, vector similarity search, safety and moderation signals, and paid or promoted placement, as described in Section 4.3 of the Terms |
| Matching insights (including AI tags, AI insights, and AI breakdowns on profiles) | Short descriptions or labels that summarize why a profile was suggested or what you may have in common | Generated from your and the other member's answers and our ranking results, using pre-written templates and rules |
| ReMatch AI | A paid feature that suggests a curated set of profiles and describes what you share | Candidates within your age and distance range are selected using vector similarity and rule-based scoring of shared questionnaire answers (weighted by category) and of patterns in your recent activity. The explanation text and tags are generated from pre-written templates. Scores are relative and may be adjusted for display. |
| Match Meter | A compatibility score and a view of answers you and another member have in common | Automated comparison of questionnaire answers by category using scoring rules. It may show the other member's answers alongside yours. |
| Astrology insights (including Astro Match, kundli, and Western compatibility) | Charts, compatibility scores, dosha checks, readings, and traditional remedies | Astrological calculations from birth date, time, and place using astronomical ephemeris data and traditional rules, with readings produced by rule-based and template text and, where enabled, a language model that we host. Entertainment only (Terms Section 10). |
| Content moderation | Helps detect profiles, photos, and messages that may violate the Agreement or the law | Profile text, including names and free-text fields, may be reviewed by a large language model. Messages and other text may be screened by machine-learning models we host and by rule-based filters (for example, for restricted terms, contact details, scams, and child safety signals). Photos may be reviewed by Amazon Rekognition and compared with images on the web using Google Cloud Vision. Human reviewers make or review many decisions. |
| Verification and age assurance | Photo verification, liveness checks, photo-to-selfie comparison, duplicate-account detection, public figure detection, and age estimation | Computer vision and face analysis by Amazon Rekognition, EveryPixel, models we run ourselves, and an on-device face detection model, as described in our Biometric Data Policy |
| Fraud, spam, and abuse detection | Helps detect bots, fake or duplicate accounts, scams, and banned members | Automated risk signals such as Google reCAPTCHA scores, app and device integrity checks, device and account identifiers, phone number checks, approximate location from IP address, duplicate-face signals, rate limits, and rules |
4. YOU ARE TOLD WHEN YOU ARE CHATTING WITH AI
4.1 Persistent label. The AI Companion displays a label at the top of the conversation, "You are chatting with AI, not a person." The label stays visible throughout every conversation.
4.2 Separate conversations. The AI Companion appears as its own conversation, separate from your conversations with members. It does not send messages to other members on your behalf and does not pose as a member.
4.3 Other automated messages. Messages in the Second Innings Official conversation come from Second Innings and may be automated, for example service notices, offers, and surveys. Messages from members are written by those members, and we do not use AI to write messages to members in your name.
5. LIMITS AND MISTAKES
5.1 AI can be wrong. AI outputs are generated automatically and may be inaccurate, incomplete, outdated, biased, inappropriate, offensive, repetitive, or unsuitable for your situation, and may conflict with other outputs (Terms Section 9.3). The AI Companion may sound confident when it is wrong. It does not know you, your matches, or your full circumstances, and it cannot verify anything a member tells you.
5.2 Scores and insights are estimates. Compatibility scores, percentages, meters, tags, and insights are relative estimates produced by our systems from information members provide. They are not measurements, predictions, or guarantees, may be adjusted for display, and may differ between features or over time (Terms Section 10.2). A high score does not mean a person is safe or honest, and a low score does not mean you are incompatible.
5.3 Automated tools miss things. Moderation, verification, age assurance, and fraud tools can make mistakes in both directions. They may fail to detect harmful content or members, and they may restrict permitted content or legitimate members. Do not treat the absence of a warning, or the presence of a badge, as a sign that a member is safe (Terms Sections 6.3 and 6.4).
5.4 Language. The AI Companion replies in English. Other features may support English, Hindi, and Spanish, and quality may vary by language.
5.5 Your judgment comes first. Independently evaluate any AI output before relying on it. You are responsible for your decisions and actions (Terms Section 8.4).
6. NO PROFESSIONAL ADVICE
6.1 AI Features are provided for general informational, conversational, and entertainment purposes. They do not provide, and are not a substitute for, medical, psychological, psychiatric, therapeutic, counseling, legal, financial, tax, astrological, matchmaking, or other professional advice, diagnosis, or treatment (Terms Section 8).
6.2 The AI Companion is not a licensed therapist or mental health professional and does not provide therapy, psychotherapy, or crisis counseling. It is not designed or intended for anyone in crisis. If you need support, contact a qualified professional or the crisis resources listed in Section 9.
7. YOUR DATA AND AI FEATURES
7.1 AI Companion conversations. Your AI Companion conversation history is stored only on your device, not on our servers. To generate a reply, each message you send is transmitted through our servers to our large language model provider, and the reply is returned to your device. We do not save your AI Companion messages or replies to your account. We keep counts of the AI messages you use, so that we can apply daily limits, and counts of crisis responses without any content (Section 9.8). Our systems may also keep limited technical and security logs as described in our Privacy Policy. You can clear your AI Companion history on your device in the app, and signing out removes locally stored messages.
7.2 Model providers. Our large language model providers process Inputs and outputs on our behalf under contracts with us and are not permitted to use them to train their models. Providers may keep data for limited periods under their terms, for example for abuse monitoring, security, or legal compliance.
7.3 No training of generalized models. We do not sell Inputs, and we do not use your private messages with other members, your Inputs, or your sensitive personal information to train generalized AI models that are made available to third parties (Terms Section 9.6). We may use aggregated or de-identified information, and information about how features perform, to evaluate, maintain, and improve the Services.
7.4 Moderation records. When content is reviewed by automated moderation, we may keep the content reviewed, the automated result, and related decisions, for safety, audit, appeals, and legal purposes, as described in our Privacy Policy.
7.5 What to avoid sharing. Do not share with the AI Companion another person's personal information or private messages without their permission, or information you do not want processed as described here, such as financial account details, passwords, or one-time codes (Terms Section 9.5).
7.6 Biometric Data. Verification and age assurance features use Biometric Data only as described in our Biometric Data Policy, with your consent.
8. AUTOMATED DECISIONS AND HUMAN REVIEW
8.1 Automated systems help decide which profiles are shown and in what order (Terms Section 4.3), and help detect content and accounts that may violate the Agreement. Depending on the result, an automated system may hide a profile until it is reviewed, hold, delay, or not deliver a message, require additional verification, or limit features or visibility (Terms Section 6.4).
8.2 If you believe an automated decision about your account or content is wrong, you can ask for human review by contacting support@secondinnings.app. Where applicable law gives you the right not to be subject to a decision based solely on automated processing that significantly affects you, or to obtain human review, express your point of view, and contest the decision, you can exercise that right the same way (Terms Section 6.6).
9. CRISIS SAFETY PROTOCOL
This Section describes the protocol we maintain for the AI Companion to respond to expressions of suicidal ideation, suicide, or self-harm, and to prevent the AI Companion from producing content about suicidal ideation, suicide, or self-harm in reply to them. We publish it to comply with California Senate Bill 243 and to inform all members.
9.1 Scope. The protocol applies to every AI Companion conversation, for all members, in all locations.
9.2 AI disclosure. The AI Companion shows the persistent "You are chatting with AI, not a person" label described in Section 4.1, and the crisis response itself states that it comes from an automated AI and not a person.
9.3 Detection. Before a message is sent to the language model, our servers check it against phrases that indicate suicidal ideation, suicide, self-harm, or crisis, in English, Spanish, Hindi, and Hinglish. Detection is intentionally broad. In addition, the language model is instructed to identify crisis messages that the phrase check does not catch, so that our fixed crisis response is shown instead of a model-generated reply. Crisis detection runs before other filters and before daily message limits are applied, and it also applies to messages that are longer than the normal length limit.
9.4 Fixed crisis response. When crisis language is detected, the AI Companion responds with a fixed message that we have written in advance, not with a reply generated by the model. The fixed message:
(a) states that it is an automated AI and not a person;
(b) refers the member to the 988 Suicide and Crisis Lifeline (call or text 988) and the Crisis Text Line (text HOME to 741741); and
(c) tells the member to call 911 in an emergency in the United States, 112 in the European Union, the United Kingdom, or India, or their local emergency number.
In the app, the response is shown as a highlighted card with buttons to call or text 988, text HOME to 741741, and call 911, and a note to use the local emergency number (such as 112) outside the United States.
9.5 The conversation does not continue as normal. When the fixed crisis response is triggered, the message is not answered by the model, and for messages caught by the phrase check the message is not sent to the model at all. A crisis response does not count against the member's daily AI message limit.
9.6 Help is always available. The AI Companion screen has a "Get help now" link that opens the same crisis resources at any time, including when a member has no AI messages remaining.
9.7 Limits of the protocol. The protocol is automated. It may not detect every expression of crisis, and it may show crisis resources when they are not needed. AI COMPANION CONVERSATIONS ARE NOT MONITORED IN REAL TIME BY PEOPLE, AND A CRISIS DETECTION DOES NOT ALERT SECOND INNINGS STAFF, EMERGENCY SERVICES, OR ANYONE ELSE. If you or someone else is in danger, contact emergency services directly. The crisis response is provided in English.
9.8 Records and annual reporting. Each time the fixed crisis response is issued, we record a count for that day and the detection method (phrase check or model). These counts do not include the member's identity, account identifier, or the content of any message. Beginning in 2027, we will use them to report to the California Office of Suicide Prevention each year as required by law, including the number of crisis service provider referral notifications issued in the preceding calendar year and a description of this protocol.
9.9 Adults only. Second Innings, including the AI Companion, is only for adults aged 18 and over (Terms Section 2). Companion chatbots may not be suitable for some minors. If we learn that a person under 18 is using the Services, we may restrict or terminate the account (Terms Section 2.3).
9.10 Review and updates. We may review and update the detection phrases, model instructions, crisis message, and resources from time to time, for example to improve detection, reflect changes in available crisis services, or meet legal requirements. Material changes will be reflected in this Section.
10. EU ARTIFICIAL INTELLIGENCE ACT TRANSPARENCY STATEMENT
10.1 This statement is provided under Article 50 of Regulation (EU) 2024/1689 (the EU Artificial Intelligence Act) for members in the European Union and the European Economic Area.
10.2 Interacting with AI. The AI Companion is an AI system that interacts directly with you. We inform you that you are interacting with an AI system through the persistent label described in Section 4.1, at the start of and throughout every conversation, and in this Disclosure.
10.3 AI-generated content. Text generated by the AI Companion appears only in the AI Companion conversation, which is labeled as AI. Astrology readings and matching insights are presented as automated insights and are described in Section 3.
10.4 Face analysis. Our verification and age assurance features analyze images of faces and may estimate characteristics such as an approximate age range. We inform you of this here and in our Biometric Data Policy, and we process this data only with your explicit consent. These features are not used to categorize members by race, ethnicity, religion, political opinions, trade union membership, sex life, or sexual orientation.
10.5 Human oversight. Members can request human review of moderation and verification decisions (Section 8.2).
10.6 Contact. Questions about our use of AI may be sent to legal@secondinnings.app, which is also our single point of contact for authorities and members under the EU Digital Services Act (Terms Section 27.1(d)).
11. OTHER LAWS
Some jurisdictions have specific rules for AI systems, companion chatbots, and AI in mental health services. We design the AI Companion to provide clear AI disclosure, crisis referrals, and no therapy or professional services, as described in this Disclosure. Nothing in this Disclosure limits any right you have under applicable law that cannot be limited by contract.
12. CHANGES TO THIS DISCLOSURE
We may update this Disclosure as our AI Features, providers, and the law change. We will post the updated version at secondinnings.app/ai-disclosure and change the "Last Updated" date, and we will notify you of material changes as described in Section 28 of the Terms.
13. CONTACT
- General questions and requests for human review: support@secondinnings.app
- Legal questions, authorities, and AI Act or Digital Services Act matters: legal@secondinnings.app
- Privacy requests and India Grievance Officer (Amit Pandya): privacy@secondinnings.app
Second Innings, Inc., 9206 Sunshine Pl, Downey, CA 90240, USA