> For the complete documentation index, see [llms.txt](https://docs.visionmakers.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.visionmakers.io/legal/ai-transparency.md).

# AI Transparency (EU AI Act)

Version 2.0 · Effective 24 October 2026

This statement explains how **Vision Makers Technologies Inc.** uses artificial intelligence, what that means for you,
and how we approach the EU AI Act and comparable regimes.

## 1 · Where AI is used

| Feature | What the AI does |
| --- | --- |
| Conversational interface | Interprets your question and generates an answer |
| Mindshare scoring | Ranks attention using proprietary models |
| Risk interpretation | Flags liquidity, concentration and contract risk indicators |
| Wallet ranking | Scores tracked wallets by observed realised performance |
| Agents | Runs multi-step workflows and pre-stages actions |
| Summarisation | Condenses news and social sources |

**You are always interacting with a machine.** AI-generated output is labelled as such in the
interface.

## 2 · Classification under the EU AI Act

Our assessment is that the platform's AI systems are **limited-risk** and subject principally to
**transparency obligations**: users must know they are interacting with AI and that content is
machine-generated.

We consider that these systems are **not**:

- prohibited practices under Article 5 (no subliminal manipulation, no social scoring, no exploitation of vulnerability);
- high-risk under Annex III, in particular, we do **not** perform creditworthiness assessment or credit scoring of natural persons, and we do not use AI for employment, education, essential services, law enforcement or biometric categorisation.

<div class="callout">
<span class="callout-label">Confirm before relying on this</span>
The high-risk boundary for financial-services AI is fact-specific, and the analysis changes if the
platform ever offers credit, insurance-like products, or automated advice to EU consumers.
<strong>This classification must be confirmed by counsel</strong> and reassessed whenever a
materially new AI capability ships. GPAI obligations fall on our model providers; our obligations
as a deployer are addressed here.
</div>

## 3 · Transparency commitments

1. **Disclosure**: AI-generated content is identified as machine-generated.
2. **Grounding**: answers are generated from retrieved platform data, not model memory. Where data is unavailable the system says so rather than interpolating.
3. **Evidence**: interpretations state the inputs behind them, so a conclusion can be checked rather than merely trusted.
4. **No opaque composite scores**: we do not publish a single uninterrogable "safety score"; components are exposed.
5. **Limitations stated**: outputs carry the caveat that they may be incomplete or wrong and are not advice.

## 4 · Human oversight

AI output is **advisory**. It never executes a transaction. Every action requiring value movement
terminates at your signature, the human-in-the-loop control is architectural, not procedural.

## 5 · Automated decision-making about you

Two processes may produce automated decisions with a significant effect:

| Process | Effect | Your right |
| --- | --- | --- |
| **Automated KYC verification** | May reject or delay verification | Human review on request |
| **Sanctions / risk screening** | May restrict access | Human review on request, to the extent legally permitted |

To request human review, contact [admin@visionmakers.io](mailto:admin@visionmakers.io). We will have a person re-examine the
decision, explain the outcome so far as law allows, and correct it where wrong. In sanctions
matters our ability to explain may be legally constrained.

## 6 · Data used by AI

Prompts and interaction context may be processed by third-party model providers listed in
[Sub-processors](/legal/sub-processors). We do not sell prompts, and we contractually require
that content is **not used to train providers' general-purpose models** where that commitment is
offered. Aggregated, de-identified interaction data may improve our own scoring models. See
[Privacy Policy §5](/legal/privacy).

## 7 · Accuracy, bias and testing

Models are evaluated for accuracy and failure modes before release and monitored afterwards.
Known limitations:

- **Social signal is adversarial**: participants actively manufacture the appearance of attention, and scoring can be gamed.
- **Coverage is uneven**: newer or smaller assets have thinner data and less reliable interpretation.
- **Training data reflects historical market behaviour**, which may not hold.

We publish material limitations rather than concealing them, and we correct errors when found.

## 8 · AI literacy

Staff involved in operating or deploying AI systems receive training appropriate to their role,
consistent with Article 4 of the EU AI Act.

## 9 · Reporting a problem

Report inaccurate, harmful or unexpected AI output to [admin@visionmakers.io](mailto:admin@visionmakers.io). Include the
prompt, the output and the time, so we can reproduce it.

## Contact

[admin@visionmakers.io](mailto:admin@visionmakers.io)