✦ Effective April 2026
Last updated: April 24, 2026
Diraflow AI is built on the belief that high-quality AI requires not just technical excellence, but ethical responsibility. This Code of Conduct defines the standards we hold ourselves, our team, and our partners to — without exception.
1. Our Core Commitments
Everything we do at Diraflow is guided by four principles:
- Integrity — We are honest about what we can deliver, how we work, and the limitations of our data.
- Responsibility — We take ownership of the downstream impact of the training data we produce.
- Respect — We treat every person we work with — clients, annotators, and partners — with dignity and fairness.
- Transparency — We communicate openly about our processes, pricing, and any issues that arise.
2. Commitment to Clients
When you work with Diraflow, you can expect:
- Honest scoping — we will not overstate our capabilities or underestimate complexity to win a contract.
- Clear communication — we will keep you informed at every stage of a project, including when things are not going to plan.
- Data confidentiality — all client data, project details, and proprietary information are handled with strict confidentiality and never shared with third parties without explicit consent.
- Quality accountability — if delivered work does not meet agreed specifications, we will remediate it at no additional cost.
- No hidden fees — pricing is agreed upfront and in writing. Changes to scope are discussed before work begins.
3. Commitment to Annotators & Contributors
Our annotators and human contributors are the foundation of everything we build. We commit to:
- Fair and timely compensation — contributors are paid fairly for their work and on schedule, always.
- Clear task instructions — we provide unambiguous guidelines and calibration materials so contributors are set up to succeed.
- Safe working conditions — we do not expose contributors to harmful, distressing, or exploitative content without appropriate safeguards, support, and informed consent.
- Feedback and respect — we treat contributor feedback as valuable signal, not noise. Contributors can raise concerns without fear of retaliation.
- No surveillance overreach — monitoring of contributor work is limited to what is necessary for quality assurance and is never used punitively.
4. Responsible AI Practices
We recognise that training data shapes AI behaviour at scale. We take that responsibility seriously:
- We do not knowingly produce training data designed to deceive, manipulate, or harm end users.
- We do not produce datasets intended to train systems for illegal surveillance, targeted harassment, or the erosion of individual rights.
- We apply extra scrutiny to safety-critical domains including healthcare, legal, financial, and security applications.
- We maintain internal review processes for projects that carry elevated ethical risk, and reserve the right to decline engagements that conflict with these values.
- We stay informed about evolving AI safety norms and update our practices accordingly.
5. Conflicts of Interest
Diraflow team members are expected to avoid situations where personal interests conflict with the interests of clients or contributors. This includes:
- Disclosing any financial or personal relationships with client organisations before a project begins.
- Not accepting gifts, payments, or favours from clients, partners, or vendors that could influence business decisions.
- Not using client data or project insights for any purpose beyond the scope of the engagement.
6. Diversity, Equity & Inclusion
We believe the best AI data reflects the full diversity of human experience. We actively work toward this by:
- Recruiting annotators and contributors from diverse linguistic, cultural, and geographic backgrounds.
- Designing annotation tasks that do not systematically disadvantage or misrepresent any group.
- Maintaining a zero-tolerance policy for discrimination, harassment, or exclusion within our team and contributor community — based on race, gender, age, religion, disability, sexual orientation, or any other characteristic.
7. Environmental Responsibility
We acknowledge that AI development has an environmental footprint. We commit to:
- Minimising unnecessary compute usage in our internal tooling and workflows.
- Preferring cloud providers and infrastructure with credible sustainability commitments where feasible.
- Being honest with clients about the environmental costs of large-scale data and model training work.
8. Reporting a Concern
If you believe Diraflow has acted in violation of this Code of Conduct — whether as a client, contributor, partner, or member of the public — we want to hear from you.
Reports can be made confidentially and will be reviewed by senior leadership within five business days. We do not tolerate retaliation against anyone who raises a genuine concern in good faith.
9. Updates to this Document
This Code of Conduct is a living document. We review and update it as our work evolves, as AI norms develop, and as we learn from the communities we work with. The effective date at the top of this page reflects the most recent revision. Material changes will be communicated to active clients and contributors directly.
Diraflow