AI Disclaimer & AI Usage Policy

1. Purpose of This Document

This document explains the limits of Nerivo AI’s AI features, operated by Noveriks Team, and the way we expect them to be used. It forms part of our Terms & Conditions.

It is written to be useful rather than alarming. Used well, these features are genuinely helpful for drafting, learning, exploring ideas, writing code, and finding your way to sources. Used as an unquestioned authority, they can mislead you.

2. How AI Output Is Produced

When you send a message, it is passed to a third-party AI model together with the conversation context. The model generates a response by predicting likely text based on patterns in its training, and, where a search feature is used, on information retrieved at that moment.

This means output is generated rather than looked up. A model does not have a database of guaranteed facts it consults before answering, and it will often produce a fluent, confident answer even where it is mistaken.

Our Deep Thinking mode asks the model to reason further before answering. This can improve difficult answers, but it does not change the fundamental point: a longer, more careful answer can still be wrong.

3. AI Output Can Be Wrong

AI output may be inaccurate, incomplete, outdated, biased, or entirely fabricated. It may cite sources that do not exist, or attribute statements to sources that do not contain them. It may be confidently mistaken about a version number, a date, a statistic, a legal requirement, or a piece of code.

AI output can be inconsistent: asking the same question twice may produce different answers. A single answer is not a considered opinion.

4. Review Before You Rely

Before relying on AI output, we ask you to:

• check anything factual against a reliable, independent source
• treat specific figures, dates, citations, and legal or medical statements as unverified until you have confirmed them
• test code before you run it on anything that matters
• keep a copy of anything important that you have generated
• consider what happens if the output is wrong, and whether that risk is acceptable to you

This applies most strongly to output you intend to publish, submit, act on, or build upon.

5. Not Professional Advice

AI output is not professional advice and does not create a professional relationship of any kind. It is not a substitute for advice from a qualified professional in any field.

Be especially careful in high-risk areas:

• Medical and health decisions — consult a qualified health professional
• Legal matters — consult a qualified lawyer in the relevant jurisdiction
• Financial, tax, and investment decisions — consult a qualified adviser
• Safety-critical engineering, construction, and industrial tasks — obtain the required qualified sign-off
• Employment, immigration, and regulatory compliance — confirm with a qualified professional

Where a decision could cause harm to a person or significant financial or legal loss, treat AI output as a starting point for discussion with a professional, never as the answer.

6. You Remain Responsible

You decide what to do with what the Service produces. You are responsible for how you use it, for anything you publish or send based on it, for any code you execute, and for any consequence that follows.

You should also check that your use complies with the rules that apply to you — including the terms of any service, employer, or institution you are working within, and any professional obligations you hold.

7. Deep Thinking Mode

Deep Thinking asks the model to spend more effort on a task before answering. It is useful for problems that benefit from step-by-step reasoning, such as planning, analysis, and complex code.

It is not a correctness guarantee. Extended reasoning can produce a more detailed answer that is still wrong, and a plausible chain of reasoning can rest on an incorrect premise. Review the conclusion as carefully as the working.

8. Nerivo Reacher (Deep Research)

Deep Research differs from ordinary chat because it retrieves live information from the web. That makes it more current and more checkable, but it introduces its own limits.

What it does

• interprets your research request
• generates search queries
• retrieves publicly accessible pages from the web
• extracts information and evidence from those pages
• compares findings against your requirements
• summarises results and provides source links

What limits apply

• Coverage is not the whole web. Only sources reachable at the time of the run can be used.
• Results may miss relevant material, particularly where the best sources sit behind logins, block automated access, or are not indexed.
• Search availability and behaviour change over time and are outside our control.
• A retrieved page may itself be wrong, out of date, promotional, or deliberately misleading.
• Findings are assembled by automated systems and may be misread or misattributed.
• Availability limits may mean a run is shorter or narrower than you intended.

We do not describe a result as verified unless a verification step actually occurred. Where a research result is reported as unverified, unsupported, or unmatched, that reflects the run honestly and should not be read as a claim that no information exists on the subject.

Where a research step fails — for example, because a search step is unavailable or a limit was reached — the Service reports that failure rather than presenting it as an absence of information. A failure is not a finding.

9. External Websites

Deep Research retrieves content from websites we do not control and do not endorse.

• Their content, accuracy, and availability are the publisher’s responsibility, not ours.
• Their privacy practices apply to the request they received, and are outside our control.
• A source appearing in a result means it was retrieved and read — not that we vouch for it.
• Links may change, expire, or lead somewhere different from what was retrieved.

Treat content retrieved from the open web the way you would treat any other unverified online material, and check the source before relying on a claim.

10. Prices, Availability, and Time-Sensitive Information

Prices, stock levels, product specifications, delivery times, opening hours, availability of a service, and published terms change frequently and often without notice. Information of this kind may be out of date the moment it is retrieved, and may already have been wrong at the source.

Always confirm time-sensitive information directly with the vendor, provider, or official source before acting on it. A research result is not a quotation, a contract, or a guarantee of availability.

Likewise, AI models have knowledge cut-offs. Where a question depends on recent events and no live search is used, the answer may simply not know about something that has since happened.

11. Images, Documents, and Extraction

When you attach an image or document for analysis, the Service may describe it, read text from it, or extract structured information. Extraction is not perfect:

• handwriting, low-resolution images, unusual fonts, and complex layouts may be misread
• numbers, units, and identifiers are particularly easy to misread
• a confident extraction is not a guarantee of accuracy

Where an extracted value matters — an amount, an ID number, a dosage, a date — confirm it against the original document.

12. Generated and Executed Code

AI-generated code may contain bugs, security vulnerabilities, or dependencies that are unsuitable or unavailable. It may be plausible and still not do what it claims.

Review and test generated code before using it. Do not run generated code against production systems, sensitive data, or anything you cannot afford to lose without reviewing it first.

Where the Service offers a sandbox to execute code, the environment is limited and may be reset without notice. It is intended for testing and learning, not for anything you need to persist. You execute code at your own risk.

13. Voice Features and Transcription

Voice input is converted to text by automated speech recognition, which makes mistakes — particularly with accents, background noise, technical vocabulary, and proper nouns. A transcription error can change the meaning of a request entirely.

Spoken responses are generated by synthetic speech and may mispronounce names, numbers, or technical terms. Where exact wording matters, use text.

14. A Privacy Reminder

Prompts are transmitted to third-party AI providers to be answered. Do not enter information you would not be comfortable sending to a third-party processor, including:

• credentials, passwords, or API keys
• other people’s personal or sensitive data
• confidential business information you are not permitted to disclose
• anything you are contractually or legally restricted from sharing

See our Privacy Policy for details of how prompts, responses, and attachments are handled.

15. Reporting Problematic Output

If output is harmful, clearly wrong in a way that matters, or appears to have been produced in breach of our Acceptable Use Policy, please tell us:

• Email: help.mahinulislam@gmail.com

Reports help us improve the Service and identify patterns that need attention.

16. Changes to This Document

We may update this document as the Service evolves. Material changes will be communicated in the same way as changes to our Terms & Conditions.

17. Contact

Questions about this document, or reports of output that appears harmful or clearly wrong, can be sent to:

• Email: help.mahinulislam@gmail.com
• Legal entity: [LEGAL ENTITY NAME]

For the operator’s full legal identity and registered address, see our Terms & Conditions. For how prompts, attachments, and research requests are handled, see our Privacy Policy.