AI Use Policy

This AI Use Policy ("Policy") governs your access to and use of Orbit AI, the artificial intelligence research, writing, source discovery, and draft-review capabilities of the Service provided by OrbitSub, a sole proprietorship owned and operated by Karan in New Delhi, India, and registered as a micro enterprise under the Udyam (MSME) scheme of the Government of India ("OrbitSub", "we", "us", or "our"). This Policy forms an integral part of our Terms of Service, Privacy Policy, and Academic & Research Disclaimer. Capitalised terms used but not defined herein have the meanings given in the Terms of Service.

1. What Orbit AI Is & Capabilities Covered

1.1 Architecture. Orbit AI is an integrated suite of research and academic writing workflows powered by third-party AI model providers (as detailed in our Privacy Policy) combined with real-time retrieval from open scholarly bibliographic databases (including OpenAlex, Crossref, Semantic Scholar, PubMed, arXiv, Unpaywall, and ORCID) and live web search indices.

1.2 Covered Features. This Policy applies to all Orbit AI features across the Service, including:

1.3 Text-Only Modality. Orbit AI generates and analyses text, structured data, mathematical notation (LaTeX), and bibliographic metadata (BibTeX / RIS / CSL). It does not generate synthetic images, audio, or video. Orbit AI is software—not a human researcher, licensed professional, or peer reviewer—and does not possess independent credentials or legal accountability.

2. Known Technical Limitations & What Orbit AI Cannot Promise

2.1 Hallucinations & Inaccuracies. Generative AI systems predict text probabilistically and can make mistakes. Orbit AI may produce statements that are factually incorrect, incomplete, outdated, methodologically flawed, or biased, and may present inaccurate statements in confident, scholarly language.

2.2 Source Attribution & Citation Limitations. Even when Orbit AI retrieves real DOIs, PMIDs, or arXiv identifiers from scholarly indices or displays excerpts from your uploaded PDFs, it may:

1. misinterpret the findings, sample size, statistical significance, or methodology of the cited paper;
2. attribute a finding to the wrong author or publication;
3. cite a secondary review article rather than the primary empirical study; or
4. fail to identify that a paper has subsequently been corrected, criticised, or retracted.

The presence of a formatted citation or a green/verified badge in a Citation Report confirms bibliographic resolution against external metadata registries, not that the cited work substantively proves your specific sentence.

2.3 No Professional, Clinical, or Safety-Critical Advice. As set out in our Academic & Research Disclaimer, Outputs do not constitute medical, clinical, legal, financial, tax, engineering, or safety-critical advice and must never be relied upon for patient care, clinical trials, legal representation, or hazardous laboratory operations.

2.4 No Guarantee of Uniqueness or Plagiarism Clearance. Independent users submitting similar research questions or referencing the same canonical papers may receive similar Outputs. Orbit AI does not warrant that Outputs are free from inadvertent textual similarity to existing literature.

3. Your Responsibilities & Academic Integrity

When using Orbit AI, you agree to observe the following five mandatory principles:

1. Verify Every Fact, Quote, and Citation: You must personally inspect the original primary source for every factual assertion, statistical figure, direct quotation, and bibliographic reference before relying on, exporting, publishing, or submitting your work.
2. Maintain Human Authorship & Editorial Control: You are the sole author and editor of your work. You must review every proposed edit, inline suggestion, and generated section before incorporating it into your manuscript.
3. Comply with Institutional, Publisher, and Funder Policies: Universities, examination boards, thesis committees, academic journals (including COPE, ICMJE, IEEE, ACM, Springer Nature, and Elsevier guidelines), and grant agencies maintain specific rules regarding whether and how generative AI tools may be used and disclosed. You are solely responsible for knowing and complying with the rules applicable to your submission, including including an explicit AI Disclosure / Methods Statement where required.
4. Accountability for Published Content: Any Content you publish on OrbitSub or export for external submission is your Content under our Terms of Service, regardless of whether Orbit AI assisted in drafting or reviewing it.
5. Lawful Input Rights: You must only upload PDFs, datasets, manuscripts, and prompts that you have the lawful right to process and that do not violate third-party intellectual property, confidentiality, or data protection rights.

4. Strictly Prohibited Uses of Orbit AI

You must not use Orbit AI, directly or indirectly, to:

1. Academic Fraud & Research Misconduct: Fabricate empirical data, falsify experimental results or statistical p-values, invent non-existent references, generate fraudulent peer reviews, or operate commercial contract-cheating ("essay mill" or "paper mill") services where work is produced for third parties to submit under false authorship;
2. Unlawful & Harmful Content: Generate, promote, or disseminate any Content prohibited by Clause 9 of our Terms of Service or our Community Guidelines;
3. CSAM & Exploitation: Generate, request, or process any material that sexualises, exploits, or harms minors in any manner;
4. Harassment, Defamation & Hate Speech: Generate targeted harassment, defamatory statements, doxxing material, or hate speech attacking protected groups;
5. Disinformation & Impersonation: Generate coordinated disinformation campaigns, fraudulent academic credentials, or material that impersonates real researchers, institutions, journals, or public officials;
6. CBRN, Weapons & Dangerous Materials: Seek or generate actionable instructions for synthesising chemical, biological, radiological, or nuclear (CBRN) agents, explosives, firearms, illicit narcotics, or hazardous toxins;
7. Self-Harm: Generate content that encourages, instructs, or glorifies suicide, self-injury, or eating disorders;
8. Cyberattacks & Malware: Write malicious code, exploit payloads, ransomware, phishing templates, or instructions for compromising computer networks;
9. Automated High-Risk Decisions: Use Outputs to make automated decisions affecting individuals' healthcare, clinical diagnosis, legal rights, employment, academic admission/grading, creditworthiness, housing, or access to essential services without qualified human review;
10. Unlawful Personal Data Processing: Input special-category personal data, unconsented patient health records (PHI), or confidential personally identifiable information (PII) of third parties without a valid legal basis;
11. Safety Bypass & Prompt Injection: Attempt to bypass, jailbreak, override, or manipulate Orbit AI's safety filters, system guardrails, or rate limits, or attempt to extract internal system prompts, instructions, or orchestration configurations;
12. Model Extraction & Competing AI Training: Use Outputs, synthetic datasets, or automated queries from Orbit AI to train, distill, fine-tune, benchmark, or evaluate competing machine learning or large language models;
13. Unauthorised Automation: Access Orbit AI through unauthorised bots, scrapers, browser automation, or shared account credentials.

5. Automated Guardrails, Transparency & Enforcement

5.1 Safety Filters. Both OrbitSub and our AI model providers employ automated input and output safety classifiers. If a prompt or uploaded passage triggers safety rules regarding severe harm, unlawful conduct, or prompt injection, Orbit AI may refuse the request, truncate the response, or redact the unsafe output.

5.2 AI Transparency Labelling. Within the OrbitSub workspace, Orbit AI interactions, proposed edit cards, and automated review runs are clearly distinguished from human collaborator comments and Live Room edits in accordance with international AI transparency standards (including Article 50 of the EU Artificial Intelligence Act).

5.3 Enforcement. Repeated or severe violations of this Policy may result in temporary throttling of Orbit AI access, suspension of AI features, or permanent Account termination in accordance with our Content Moderation Policy.

6. How Orbit AI Handles Your Data

6.1 Strict Zero-Training Guarantee. We do not use your research manuscripts, uploaded PDFs, Reference Library items, Orbit AI prompts, or generated Outputs to train, fine-tune, or improve any foundation AI models. We contract with our third-party AI model providers strictly via enterprise API agreements that prohibit them from training their models on customer API payloads.

6.2 Data Transmission During Requests. When you invoke an Orbit AI feature:

• Your prompt, the relevant contextual excerpt of your active draft, and selected excerpts from your attached Reference Library or uploaded PDFs are transmitted over TLS 1.2+ to our AI model providers solely to generate the requested Output; and
• Where literature or web retrieval is enabled, search queries derived from your research question or claim are sent to scholarly metadata registries and web search providers.

Full details of our sub-processors and zero-training safeguards are set out in our Privacy Policy and Data Retention Policy.

6.3 Managing Your AI History. Your Orbit AI conversation threads (AiConversation), messages (AiMessage), review runs (AiReviewRun), and custom prompts (AiPrompt) are stored privately in your Account so you can resume your research. You may delete individual conversations or custom prompts at any time inside the workspace, or erase all AI history permanently by deleting your Account under our Data Deletion Policy.

7. Ownership of Outputs

As between you and OrbitSub, and to the maximum extent permitted by applicable law, you retain all rights, title, and interest in the Outputs generated for you by Orbit AI, as set out in Clause 8.6 of our Terms of Service. Whether AI-assisted text qualifies for copyright protection in a given jurisdiction depends on applicable copyright law and the degree of human creative selection, arrangement, and editing you contribute.

8. Plan Quotas & Fair Use

Access to Orbit AI features is governed by the monthly usage allowances of your subscription tier (Free, Researcher, or Pro), as published on our Pricing page and tracked in real time under Settings → Billing & Usage. Fair-use rate limits also apply to protect platform stability and prevent automated abuse.

9. Reporting Inaccurate or Unsafe AI Outputs

If Orbit AI generates an Output that is harmful, unlawful, biased, or dangerously inaccurate, please report it to us at team@orbitsub.com with the subject line Orbit AI Feedback, including the prompt used and the Output received.

This Policy is effective from 1 October 2026 and was last updated on 25 October 2026.