DBR77 AI Policy

Effective date: 2026-03-25

1. Scope

This policy applies to DBR77 AI-assisted functionality across our products and services, including recommendation engines, matching and ranking tools, anomaly interpretation, narrative or report generation, scorecards and assessments, industrial reasoning features, and generative user interfaces.

2. Core principles

DBR77's AI position is:

• assistive, not autonomous,
• human-reviewed, not self-executing,
• contract-bounded, not marketing-expanded,
• data-controlled, not training-by-default.

3. Intended use

DBR77 AI features are designed to help users organize information, generate drafts, prioritize options, summarize inputs, surface patterns, and support human decision-making. Unless a product schedule expressly states otherwise, AI outputs are intended to inform professional judgment rather than replace it.

4. No-training default

Unless DBR77 and the relevant customer expressly agree otherwise in writing, customer-submitted data, customer-specific prompts, uploads, workflow records, and operational records are not used to train DBR77's proprietary models or third-party foundation models.

5. Human review requirement

AI outputs must be reviewed by appropriately qualified human personnel before they are used for:

• operational decisions,
• engineering or process changes,
• safety-related actions,
• legal or regulatory reliance,
• employment decisions,
• procurement or investment commitments,
• financial approvals,
• board-facing or customer-facing final conclusions.

6. What AI outputs are not

Unless expressly stated in a signed agreement, AI outputs are not:

• legal advice,
• tax advice,
• accounting advice,
• engineering certification,
• safety certification,
• medical advice,
• employment or HR advice,
• guaranteed ROI analysis,
• independent expert opinion.

7. Customer responsibilities

The customer remains responsible for:

• verifying the legality, quality, and sufficiency of input data,
• reviewing output accuracy, completeness, and feasibility,
• determining whether an output is suitable for the customer's context,
• obtaining any internal or external approvals required before acting on an output,
• ensuring that use of AI features complies with law, contract, and internal policy.

8. Prohibited or restricted uses

Unless separately validated, risk-assessed, and expressly contracted, DBR77 AI must not be used as the sole basis for:

• machine control or safety control,
• emergency response decisions,
• medical diagnosis or treatment,
• employment or candidate screening decisions,
• legally determinative automated profiling,
• sanctions screening or export-control clearance,
• autonomous procurement, financing, or investment decisions.

Users may not use DBR77 AI features to generate or distribute unlawful, infringing, deceptive, harassing, malicious, or privacy-violating content.

9. Model composition and routing

DBR77 may use a combination of proprietary models, third-party models, rule-based logic, retrieval systems, vector search, ranking systems, workflow automation, and human review layers. DBR77 may change the composition of these systems over time for performance, safety, cost, or operational reasons.

DBR77 is not required to disclose implementation details beyond what is appropriate for security, contractual, diligence, or regulatory purposes.

10. Product-specific overrides

Product schedules may impose stricter or more specific rules for AI features, including:

• output warnings,
• deployment constraints,
• industry-specific restrictions,
• private deployment options,
• infrastructure or residency boundaries,
• additional customer responsibilities.

If a product-specific document conflicts with this policy, the stricter rule applies for that product.

11. Public disclosure discipline

Public AI statements should be drafted conservatively and should not disclose model names, detailed architecture, or training methods unless DBR77 expressly decides to publish that information.

12. Changes

DBR77 may update this policy from time to time. The current version will be posted with a revised effective date.