AI Services Disclosure
Effective Date: July 25, 2026
Last Updated: July 25, 2026
This disclosure explains what our AI features are, how they behave, and where their limits lie. It forms part of our Terms of Service.
We would rather set expectations honestly than have you discover these limits in production.
Index
- What the Services Do
- Outputs Are Probabilistic
- Accuracy and Hallucination
- You Must Review Outputs
- Outputs Are Not Unique
- Bias
- Safety Filters Are Imperfect
- The Service Changes Continuously
- Experimental Features
- Not Professional Advice
- Synthetic Media
- Questions
1. What the Services Do
Our Services use machine learning models, operated by us and by third-party providers, to generate images, video, audio, speech, and avatars from prompts and reference material you supply.
The models predict plausible output. They do not understand your intent, verify facts, or reason about consequences.
2. Outputs Are Probabilistic
Generation is not deterministic.
- The same prompt submitted twice may produce different results.
- Small changes to a prompt can produce large changes in output.
- Quality varies between subjects, styles, and models.
- Some generations fail or produce unusable results.
Where a model supports a seed parameter, reusing a seed improves consistency but does not guarantee identical output across model versions.
3. Accuracy and Hallucination
AI-generated content may be wrong, and may be confidently wrong.
Outputs may:
- Contain factual errors or invented details
- Render text incorrectly, particularly in images
- Produce anatomically implausible results, especially hands, eyes, and fine detail
- Misrepresent a subject, place, brand, or event
- Produce audio with mispronunciations or artefacts
- Fail to follow parts of your prompt
We make no guarantee of factual correctness, completeness, or fitness for any purpose.
4. You Must Review Outputs
You are responsible for reviewing every Output before you publish, distribute, or act on it.
This matters most where an Output will be seen by the public, depicts a real person, makes a factual claim, is used in advertising or a regulated context, or forms part of a decision affecting someone.
We provide the tool. What you do with the result is yours.
5. Outputs Are Not Unique
Models generate from learned patterns. Different users submitting similar prompts may receive substantially similar Outputs.
We cannot guarantee that any Output is unique, original, or free from resemblance to existing works. Nor can we guarantee that an Output is protectable by copyright, since the legal status of AI-generated content differs by country and remains unsettled in many.
If uniqueness matters for your use case, verify it independently.
6. Bias
Models learn from large datasets that reflect the biases present in their source material.
Outputs may under-represent, stereotype, or render less accurately certain groups of people, and may reflect cultural assumptions embedded in training data. We work to reduce this, but we cannot claim to have eliminated it.
Our approach is described in our Responsible AI Policy.
7. Safety Filters Are Imperfect
We operate automated safety systems that screen prompts and Outputs.
These systems produce both false positives, blocking content that is permitted, and false negatives, allowing content that should have been blocked. A generation completing successfully is not a determination that the content is lawful or compliant with our policies.
Our Moderation Policy explains how these systems work and how to appeal.
8. The Service Changes Continuously
Models improve, change, and are retired. Providers adjust their own systems without notice to us.
As a result, output quality, style, latency, and credit cost may change over time, and a workflow that produced a particular result last month may not reproduce it today. Our commitments around model changes and deprecation are in our Model Usage Policy.
9. Experimental Features
Features labelled beta, preview, or experimental are offered for evaluation.
They may be unstable, may change substantially, may produce lower-quality results, and may be withdrawn at any time. They are provided without warranty and are not covered by any service level commitment.
10. Not Professional Advice
The Services do not provide medical, legal, financial, psychological, or other professional advice.
Do not use Outputs as a substitute for a qualified professional, and do not use the Services to make or automate decisions affecting a person's legal rights, safety, health, employment, credit, housing, insurance, or access to essential services. These restrictions are set out in our Acceptable Use Policy.
11. Synthetic Media
Content generated by the Services is synthetic. It can be mistaken for a real photograph, recording, or video.
Where you publish synthetic media, particularly media depicting a person, you are responsible for any labelling or disclosure required by applicable law. We may embed provenance metadata or visible watermarks in Outputs, and you must not remove or alter them.
The rules on depicting real people are in our Acceptable Use Policy and Biometric Data Policy.
12. Questions
If you are unsure whether the Services are appropriate for your use case, ask us before you build on them.
Email: support@inferon.ai