Blackbone develops and integrates AI systems for customers and internal use. This policy covers language models, computer vision, predictive systems, agents and retrieval-augmented applications. The EU AI Act, GDPR and other applicable rules are considered according to the system's role, risk and use.
1. Risk and responsibility
Each project identifies whether Blackbone acts as a provider, deployer or another relevant actor. Prohibited practices must not be developed or deployed. Systems in regulated high-risk contexts require the applicable risk management, data governance, documentation, human oversight and robustness measures. Obligations and their application dates are assessed for the particular system.
2. Operational principles
• Human oversight for decisions with significant effects on people.
• Clear disclosure when users interact with an AI system, where required.
• Data quality, documented sources and assessment of bias and representativeness.
• Security testing, protection against prompt injection and controlled access to tools.
• Validation of outputs and traceability appropriate to the risk.
• Clear responsibility for monitoring, correction and escalation.
3. Copyright and data sources
Training and retrieval data must have an appropriate legal basis for use. We respect applicable copyright, licensing terms and machine-readable reservations of rights. Customer confidential material is used only as agreed for the project.
4. Personal data
AI processing follows the GDPR, including data minimisation, purpose limitation, security and applicable transfer safeguards. High-risk processing may require a data protection impact assessment. Personal data should be removed or reduced before prompting where possible. Rights concerning automated decisions remain available where Article 22 applies.
5. Generated content
AI-generated or manipulated content is identified where required, particularly when realistic synthetic media could mislead users. Disclosure and provenance measures depend on the use case and applicable rules. Users should be able to understand the system's limitations.
6. Third-party and general-purpose models
Before integrating a model, the project considers the provider's terms, data handling, capabilities, limitations and applicable regulatory documentation. A model's inclusion in a project does not remove Blackbone's or the customer's responsibilities for the final system.
7. Governance and user rights
Projects should record their AI systems, owners, risk assessment and evaluation results. Users can raise concerns, request human review where applicable and seek information about processing affecting them. Referencing ISO/IEC 42001 or another framework does not imply that Blackbone holds a certificate.
8. Contact
Functional questions: contact@blackbone.ro. Personal data and rights: dpo@blackbone.ro.