Freelance Fast

Freelance Fast AI governance

EU AI Act Information

Last updated

This page explains how Freelance Fast uses AI and where the service draws boundaries to avoid automated hiring, freelancer evaluation, worker management, or other high-risk uses.

Freelance Fast uses AI only as an assistant for conversation, drafting, summarizing, translating, formatting, safety checks, and support.

Freelance Fast does not use AI to score, rank, reject, approve, shortlist, or evaluate freelancers. Project visibility is based on explicit user-selected tags, preferences, and deterministic database filters. Clients and project owners make final contact and selection decisions.

1. Operator and AI Contact

Freelance Fast is operated by ZHAR YURIY, Via Ronago, 49, 22029 Uggiate Trevano (CO), Italy. Italian tax code: ZHRYRY93D27Z138J. VAT number: 03969970130.

Questions about AI use can be sent to elchemista@gmail.com. Formal PEC requests can be sent to yuriy.zhar@ultracert.it.

2. AI Role

Freelance Fast uses OpenRouter to route AI requests to third-party AI systems and general-purpose AI models as a deployer, not as a provider of a general-purpose AI model. The currently configured models include openrouter:inclusionai/ling-2.6-flash for assistant/smart chat work and openrouter:meta-llama/llama-3.1-8b-instruct for classification, unless changed by deployment configuration. The AI features assist the service with text work and support tasks. They are not intended to make legal, financial, employment, recruitment, worker-management, or access-to-self-employment decisions.

OpenRouter provider routing and retention can depend on the selected model, downstream provider, and privacy settings. Freelance Fast does not use client conversations to train a public Freelance Fast AI model.

3. Why This Boundary Matters

The EU AI Act gives special treatment to AI systems used in employment, workers' management, and access to self-employment, including systems that filter applications or evaluate candidates. Freelance Fast avoids that use case by keeping AI output descriptive, by using deterministic visibility rules, and by leaving project, application, contact, and selection decisions to people.

4. Allowed AI Uses

  • Help a client turn a chat message into a clearer project draft.
  • Ask clarifying questions when a request is incomplete.
  • Translate, format, summarize, or rewrite user-provided text without changing meaning.
  • Summarize public GitHub information into factual technologies, repository evidence, and user-editable profile notes.
  • Assist support replies, reminders, safety explanations, and admin review.
  • Detect obviously unsafe, abusive, spammy, or legally risky requests for human review.

5. Prohibited AI Uses

Freelance Fast should not add, market, or enable any AI feature that:

  • Scores, ranks, approves, rejects, filters, or shortlists freelancers or applications.
  • Predicts freelancer quality, trust, reliability, seniority, future performance, or likely success.
  • Decides who should be contacted, hired, promoted, replaced, excluded, or deprioritized.
  • Creates hidden candidate profiles based on inferred personal traits or behavior.
  • Uses emotion recognition, biometric categorisation, or sensitive-attribute inference.
  • Uses social scoring or manipulative techniques that materially distort user behavior.
  • Markets the product as AI hiring, AI selection, AI recruitment, or AI matching.

6. Project Visibility Rules

Project visibility must stay explainable without an AI judgment. A freelancer can see or receive a project because the project and freelancer share explicit tags, selected preferences, budget range, availability, language, channel, or other deterministic database fields.

Acceptable explanations are factual, for example: "shown because the project is tagged Elixir and Telegram and your profile selected Elixir and Telegram." Avoid wording such as "AI selected you", "best match", "top freelancer", "quality score", "seniority score", or "trust score".

7. Data Minimization and Prompt Hygiene

  • Clean conversation text before it is used in AI prompts or public project drafts.
  • Do not intentionally send client contact details, phone numbers, emails, private identifiers, passwords, credentials, private repository content, or unnecessary client-identifying information to AI tools.
  • Use public GitHub information only for descriptive summaries and editable profile text.
  • Keep AI-assisted public project text under human review before publication where practical.
  • Do not use client conversations to train a public Freelance Fast AI model.

8. Transparency and Human Oversight

  • Tell users when they are chatting with the Freelance Fast AI assistant.
  • Make clear that AI drafts can be wrong, incomplete, or outdated.
  • Let clients review project text before publication or handoff where practical.
  • Let freelancers review and update their profile and contact details.
  • Let users request correction, deletion, support, or human review by email or chat.

9. Internal AI Act Checklist

  • Maintain a short AI register with feature name, purpose, provider, input data, output, and owner.
  • Train operators on allowed AI uses, privacy limits, hallucination risk, and human oversight.
  • Review new AI features before launch, especially anything touching freelancer access, project visibility, applications, selection, pricing, or work allocation.
  • Document model or provider changes and confirm prompt-minimization still works.
  • Keep deterministic visibility logic auditable and explainable by explicit tags/preferences.
  • Get legal review before adding AI ranking, AI scoring, application filtering, candidate evaluation, targeted job ads, or automated work-allocation decisions.

10. Official References

This page follows the practical boundary suggested by the EU AI Act's risk structure: transparency for direct AI interaction, AI literacy for people operating AI systems, and special caution for employment, workers' management, and access to self-employment.