Policy · Version 1.0
Acceptable use
What may and may not be built on the AxForge API and GPU services. Short version: nothing illegal, nothing the EU AI Act prohibits, and high-risk uses talk to us first.
Request high-risk review
Trust Centre
High-risk (AI Act Annex III) uses require prior review. Contact us before deploying.
Prohibited
Uses that are never permitted
| Category | What is prohibited |
|---|---|
| Illegal activity | Illegal content or activity of any kind, under applicable law. |
| Malware & intrusion | Creating or distributing malware, credential theft, or intrusion tooling for unauthorized access to systems or accounts. |
| Infrastructure abuse | Using the API or a rented machine to attack, scan or disrupt other systems or networks, to send spam, or to evade AxForge's security and metering controls. |
| Unlawful surveillance | Monitoring or tracking of people without a lawful basis. |
| CSAM | Child sexual abuse material: zero tolerance, in any form, generated or processed. |
AI Act Article 5
Prohibited AI practices
The EU AI Act bans certain AI practices outright (Article 5). They are equally banned on AxForge, in plain words:
| Practice | In plain words |
|---|---|
| Manipulative techniques | Systems that use manipulative or deceptive techniques to materially distort a person's behaviour. |
| Exploitative targeting | Systems that exploit vulnerabilities of specific people or groups, for instance because of age, disability or social situation. |
| Social scoring | Scoring or classifying people over time based on behaviour or characteristics, leading to detrimental treatment. |
| Untargeted facial scraping | Untargeted scraping of facial images from the internet or CCTV to build facial-recognition databases. |
| Emotion inference at work or school | Inferring emotions of people in workplaces or education institutions, except in the cases the AI Act itself permits. |
| Real-time remote biometric ID | Real-time remote biometric identification in publicly accessible spaces for law-enforcement purposes. |
Prior review
High-risk uses require review first
Uses that fall in the EU AI Act's high-risk areas (Annex III) require prior review by AxForge before deployment. That includes, for example:
| Annex III area (examples) | Example use |
|---|---|
| Employment | Screening, ranking or deciding on candidates and workers. |
| Credit | Credit scoring or creditworthiness assessment of people. |
| Essential services | Deciding access to essential private or public services. |
| Law enforcement | AI systems used in a law-enforcement context. |
- Planning something in these areas? Contact hello@axforge.ai before you deploy. The examples above are not the complete Annex III list. When in doubt, ask.
Your responsibilities
What stays on the customer's side
| Topic | Expectation |
|---|---|
| Special-category data | If your content includes special-category personal data (health, biometrics, beliefs and the like), you must have a lawful basis for processing it. We discourage sending it to the shared API at all. |
| Rented machines | You operate what runs on a rented machine and are responsible for its software, its data and its network behaviour during the rental. Export what you need before the rental ends. The machine and all data on it are deleted at that point. |
| Human oversight | Consequential decisions about people require human oversight. Do not wire model output directly into decisions that affect someone's rights or access. |
| Factual correctness | Model outputs are not guaranteed to be factually correct. Verify before you rely. |
| Downstream decisions | Decisions made with model output, and their consequences, are the customer's own. |
Enforcement
Suspension and appeal
Violations of this policy lead to suspension of service. If you believe a suspension is wrong, appeal to hello@axforge.ai.
Related
Nearby on AxForge
Document control
Version
Version 1.0 · Effective 2026-09-04 · Last reviewed 2026-09-04
- 2026-09-04, version 1.0: infrastructure-abuse and rented-machine responsibilities added; policy forms part of the Terms of Service.
- 2026-08-26, first published.