Human-in-the-Loop Oversight
AI systems provide assistive drafts, preliminary extractions, and comparative summaries. Final decisions (e.g., credit approvals, diligence memos, regulatory filings) require human professional judgment.
AI Governance & Safety
AI Usage & Acceptable Use Policy
Standards governing the development, deployment, and operation of generative AI and machine learning systems across Dotnitron products and forward-deployed enterprise engagements.
1. Purpose & Scope
Dotnitron Technologies Private Limited engineers AI-powered tools, systems, and products for institutional clients in banking, private equity, insurance, and professional advisory. Because our systems support high-stakes operational workflows, we adhere to strict standards of transparency, safety, and human accountability.
This Policy applies to all users, customers, personnel, and integrated third-party systems interacting with Dotnitron applications (including Underlying, Regoscope, Repdate, and bespoke customer deployments).
2. Core AI Operating Principles
Every AI pipeline engineered by Dotnitron is built on four non-negotiable architectural pillars:
Citation & Source Grounding
Outputs must be strictly grounded in verified source documents. Claims, financial figures, and compliance findings provide direct, clickable page-level citations for transparent verification.
Zero Model Training on Data
Customer confidential files, transaction workpapers, and reviewer annotations are never used to train public or foundation AI models. See our No Training Commitment.
Defense-in-Depth Prompt Safety
Multi-layered defenses against prompt injection, delimiter tampering, and data exfiltration. Untrusted inputs are isolated before execution.
3. Prohibited Use Cases
Dotnitron systems may not be used, directly or indirectly, for any of the following purposes:
- Fully Autonomous Consequential Decisioning: Deploying AI systems to make unreviewed, automated determinations that produce legal, financial, or similarly significant effects on individuals.
- Deceptive Generation or Impersonation: Generating synthetic content, deepfakes, or misleading communications designed to deceive third parties or misrepresent human involvement.
- Illegal, Discriminatory, or Harassing Processing: Processing data to generate unlawfully biased, discriminatory, defamatory, or harassing material.
- Adversarial Evasion & Jailbreaking: Intentionally testing, probing, or exploiting prompt injection vulnerabilities, bypassing safety refusal gates, or exfiltrating system instructions.
- Unconsented Processing of Special Categories of Data: Ingesting sensitive biometric, health, or children's personal data without verifiable consent and explicit authorization.
4. Customer & Reviewer Responsibilities
When utilizing Dotnitron AI systems, customers and their authorized users agree to:
- Inspect Source Evidence: Review and validate highlighted citations before incorporating AI extractions into formal workpapers or external filings.
- Ensure Data Rights: Confirm that all uploaded documents and data sources are lawfully obtained and authorized for processing under applicable privacy laws.
- Report Anomalies: Promptly report any suspected hallucinations, inaccuracies, or unexpected model behaviors to our AI governance team.
5. AI Governance Contact & Oversight
Our AI Governance Lead and Security Lead review model performance, trace metrics, and supplier updates on an ongoing basis. For questions, safety reports, or custom deployment configurations:
Dotnitron AI Governance & Ethics
Email: info@dotnitron.com
Dotnitron Technologies Private Limited · New Delhi, India