You work with AI agents
OneActor's Product Owner, architect, designer, developers, QA, security reviewer and DevOps are AI agents, not people. Every message they write in the dashboard is marked as coming from an AI agent, and so is every email an agent's work causes us to send you. No person at OneActor reads or approves their answers before you see them.
People at OneActor see your project only when you ask for support, or when we must investigate abuse, a security problem or a legal request.
What the system is for
OneActor is an AI system, provided by ONE ACTOR- SOFTWARE DEVELOPMENT, UNIPESSOAL LDA. under its own name, for planning, designing, building, testing and deploying software — web apps, phone apps for Android and iPhone, APIs, background workers and scheduled jobs — on your instructions. That is its intended purpose.
It is not intended, and may not be used, to make or support decisions about people in the areas the AI Act treats as high-risk (for example recruitment, credit, education, access to essential services), and it must not be modified or used so as to become a high-risk AI system. If you build an app that is itself an AI system, you are that app's provider under the AI Act. An app that is a high-risk AI system under Annex III of the AI Act may not be built or run on OneActor (Acceptable Use Policy).
The models behind the agents
The agents run on general-purpose AI models made by other companies — at present OpenAI's GPT models, Anthropic's Claude (through Amazon Bedrock, in the EU) and the open-weight Kimi from Moonshot. We do not train or fine-tune models. The model developers are the providers of those models under the AI Act.
The open-weight models are run for us by inference providers, never by the developers' own services in China. Which provider serves a model can change; the providers are listed on our sub-processor page.
An organization on a corporate contract may connect its own AI account: the models its owners choose then run on that account, under the organization's own contract with that provider.
Intelligence levels (Smart, Intelligent, Brilliant, Genius) choose which models and how much reasoning each agent role uses. The dashboard's "Compare levels" shows the current models per level.
What the agents can and cannot do
- They work only inside your project's workspace and through the tools we give them.
- They can read everything in your development workspace, including keys you store for development. They cannot read the keys you store for test and production.
- They cannot turn production on, and nothing is deployed until a person in your organization presses Deploy. Production must first be switched on by the organization's owner.
- They cannot submit a phone app to the App Store or Google Play: you do that yourself, in the stores' own consoles.
- They can make mistakes: misunderstand what you want, write code with bugs or security problems, claim something works when it does not. Review what they produce, test it on the test environment, and treat their output as a draft you are responsible for.
Marking of AI-generated content
- Code the agents write is not marked: the AI Act's marking duty does not cover source code (Commission guidelines C(2026) 5054, para. 68).
- Text, documents and images the agents produce for you to use outside OneActor (for example design files or copy for your site) are marked as AI-generated in a machine-readable way as far as that is technically feasible (AI Act Art. 50(2)): design files carry the mark in their metadata, and a written assessment records what cannot be marked, and why.
- Your own publications. If you publish AI-generated images, audio or video that look real (deep fakes), or AI-generated text on matters of public interest without human editorial responsibility, you must disclose it (AI Act Art. 50(4)). If your app itself chats with people or generates content, the AI Act's transparency duties for that app are yours.
Your data and the models
Content the agents work on is sent to the model provider serving the request. We use providers that do not train on it; see the Privacy Policy and the sub-processor page for where they are and how long they keep data. Content for an AI account your organization connected goes to that account's provider, under your contract with it.
AI literacy
We keep our own knowledge of these systems current (AI Act Art. 4), and our documentation explains how to work with the agents and how to check what they produce.
Questions and concerns
Write to support@oneactor.com. You may also complain to the national market surveillance authority for AI.