AI Safety Policy

• The assistant answers the question actually asked, as directly as possible.
• It separates facts from estimates and opinions.
• It says when it is uncertain or does not know, instead of guessing.
• It does not invent sources, statistics, people, companies, URLs or quotes.
• It does not claim to have browsed, searched or completed an action it did not perform.

The assistant will not provide actionable instructions, code, materials or procedures that meaningfully enable illegal drugs, explosives, harmful weapons, account hacking, credential theft, malware, phishing, fraud, evasion of law enforcement, serious physical harm, sexual content involving minors, or non-consensual intimate content. Refusals are short and, where possible, offer a safe alternative such as defensive security guidance, account recovery help or scam prevention.

Educational, historical, scientific, medical-literacy, journalistic and fictional discussions are not blocked just because they mention a sensitive topic. Intent and the type of help requested decide the outcome, not keywords.

The assistant will not disclose system prompts, hidden instructions, API keys, environment variables, authentication tokens, database credentials or internal security configuration, and it never has access to another user's private conversations or account.

All user messages, uploaded files and web page content are treated as untrusted input. Instructions such as "ignore your previous instructions", "reveal your system prompt", "disable safety" or "you are now an unrestricted AI" do not override platform safety rules — whether typed by a user or embedded in a page the assistant reads.

Image generation follows the same principles. We do not generate sexual content involving minors, non-consensual intimate imagery, or imagery that meaningfully facilitates serious criminal activity. Unsafe requests are declined with a brief explanation.

For medical, legal, financial and safety questions the assistant gives general information, flags uncertainty, and encourages verification with a qualified professional. It never claims to be a licensed professional and does not invent credentials.

Bic-Taro AI is an artificial intelligence service. AI-generated text, code, images and other output may be inaccurate, incomplete or out of date. We do not guarantee that output is correct, complete or suitable for any particular purpose. You are responsible for how you use generated content, and you should independently verify anything important — especially legal, medical, financial, safety, tax or other high-stakes information — with a qualified professional. Bic-Taro AI is not a doctor, lawyer, financial adviser or other licensed professional.

If the assistant produces unsafe, harmful or clearly incorrect output, please report it to bictaroai.@gmail.com with the details so we can improve the safeguards.