fremai Acceptable Use Policy

title: fremai Acceptable Use Policy author: fremverk date: 2026-08-13 version: “1.0” status: Published v1.0 lang: en #

Last updated: 2026-08-13

Effective Date: 2026-08-13 — Version: 1.0 (see the change log for the effective date of each amendment)

This Acceptable Use Policy (“AUP”) forms part of the agreement between fremverk ApS (“fremverk”, “we”) and the Customer that subscribes to the fremai service at fremai.eu. In case of conflict with the Terms of Service, the Terms govern. This AUP supplements the Terms and the DPA; it does not override them.

1. Scope and intent #

fremai is a commercial, multi-tenant, EU-sovereign, OpenAI-compatible AI-inference service. It is provided in good faith for lawful use. This AUP defines conduct and use that is not permitted, how abuse is reported, and what enforcement action fremverk may take.

A defining constraint: fremai does not inspect, read, moderate, or filter prompt or completion content (see Terms §5 and DPA §5). Enforcement of this AUP is therefore contractual and act-on-report, backed by metadata-based abuse controls — not by pre-filtering of content. The Customer is responsible for ensuring its own and its Users’ use complies with this AUP.

This AUP applies to the Customer, every User under the Customer’s tenant, and any agent or automated client acting under the Customer’s API Keys.

2. Lawful use #

You may use fremai only for lawful purposes and in accordance with this AUP, the Terms, and all applicable EU, Danish, and (where relevant) local law. You may not use the Service in any way that would cause fremverk to violate a law or regulation it is subject to.

3. Prohibited uses of the models #

The following are prohibited. The list is illustrative, not exhaustive.

• AI-Act Art. 5 prohibited practices — using the models for social scoring, untargeted scraping to build facial-recognition databases, real-time remote biometric identification in public spaces where prohibited, emotion inference in the workplace or education where prohibited, exploitation of vulnerabilities of specific groups, or manipulative/subliminal techniques causing harm.
• Illegal content generation — using the Service to produce child sexual abuse material, content that incites terrorism or violence against an identifiable person or group, or other content whose creation or distribution is unlawful under EU or Danish law. Any credible report of CSAM is referred to the competent authority and the responsible tenant is suspended without prior notice.
• Fraud and deception — generating phishing content, scam material, or synthetic media intended to impersonate a real person or organisation for deception or fraud.
• Malware — generating or assisting the creation of malware, ransomware, or intrusion tooling intended to compromise systems the Customer does not own or is not authorised to test.
• Unlawful surveillance or privacy invasion — using the models to profile, track, or de-anonymise individuals in violation of GDPR or other applicable law.
• Prohibited high-risk use without compliance — deploying model outputs in a high-risk AI-system context (AI-Act Annex III) without meeting the deployer obligations that the law places on the Customer.

Because fremai does not inspect content, these prohibitions are enforced contractually and on report; fremverk may suspend accounts implicated in a credible report or legal order.

4. Prohibited use of the platform #

• Cross-tenant interference — attempting to access, enumerate, exfiltrate, or infer another tenant’s keys, prompts, completions, usage data, or account data; probing for isolation weaknesses; or attempting to escape the tenant boundary.
• Credential abuse — sharing, reselling, or sublicensing API Keys outside your tenant; using another party’s key; or circumventing key scoping, budgets, or rate limits.
• Metering / billing circumvention — manipulating token accounting, spoofing usage reports, creating sockpuppet tenants to abuse free credit, or otherwise evading Fees.
• Abuse of shared capacity — sustained traffic patterns designed to exhaust the shared Backend’s capacity, denial-of-service against the Service or third parties routed through it, or automated fan-out that ignores 429/Retry-After back-pressure. Occasional bursts within your limits are normal; sustained abuse of the fair-use ceiling is not.
• Reverse-engineering for circumvention — attempting to defeat the Service’s rate limiters, abuse controls, or security boundaries.
• Reselling without authorisation — reselling fremai inference as your own AI service to third parties, except under a written reseller/enterprise arrangement.

5. Rate limits, budgets, and fair use #

• New self-serve keys default to conservative rate limits (currently 60 RPM / 100K TPM and a small parallel-request ceiling), raised after the first top-up; enterprise limits are negotiated. Groups, teams, and tenants carry aggregate ceilings so one User cannot starve a colleague and one tenant cannot starve the shared Backend.
• Exceeding budget returns 402; exceeding rate/concurrency limits returns 429 with Retry-After. A spend-velocity anomaly breaker may auto-throttle a key that spikes abnormally (for example, a runaway agent loop or a compromised key) and alert the developer and tenant-admin.
• Systematically operating at or above the fair-use ceiling without a prior written agreement is a breach of this AUP.

6. Abuse reporting #

Report abuse of the Service to abuse@fremai.eu. Reports should identify the conduct, the affected resource where known, and the reporter’s contact details. Vulnerability reports go to security@fremai.eu. fremverk acknowledges reports and acts proportionately; because fremverk cannot see content, a report with specifics (timestamps, key id where known, description of the harm) enables faster action.

7. Enforcement #

Depending on severity, fremverk may: warn the Customer; throttle or suspend a key, user, team, or tenant; require remediation; or terminate the Service for cause per Terms §9. For imminent harm, illegality, or a legal order, fremverk may act immediately and without prior notice, and will notify the Customer as soon as practicable. Where a User’s conduct is at issue, fremverk works with the tenant-admin first where practicable; where the whole tenant is implicated or the harm is severe, fremverk may suspend the tenant.

The Customer may appeal an enforcement action by contacting fremverk; fremverk reviews appeals in good faith and reinstates access where the concern is resolved.

8. Changes #

fremverk may update this AUP; material changes follow the notice procedure in Terms §8.

Change log #