GLOBAL PRIVACY & DATA GOVERNANCE POLICY
Tekbod AI treats data privacy as a technical feature, not a legal chore.
• Zero-Persistence Architecture: For PII-sensitive datasets, Tekbod AI utilizes a 'Streaming Annotation' protocol. Data is rendered in a secure browser-based sandbox and purged from the annotator’s cache immediately upon task submission.
• Sovereignty Compliance: Tekbod AI offers 'Regional Bench' options, ensuring that data never leaves the jurisdiction of origin (e.g., EU data stays with EU-based SMEs) to satisfy GDPR and the EU AI Act’s 'Human Oversight' mandates.
TEKBOD AI GLOBAL DATA GOVERNANCE FRAMEWORK (Ver. 2026.1)
1. The "Zero-Trust" Data Lifecycle
We treat client datasets as "Active Intelligence Assets" that require environmental isolation.
Ephemeral Annotation Environments (EAE): Tekbod AI utilizes volatile memory sandboxes for all high-sensitivity labeling. Data is streamed—never stored—on annotator terminals. Once a task is submitted, the local cache is cryptographically shredded (DoD 5220.22-M standard).
Protocol-Level Anonymization: Before data reaches our SME bench, Tekbod AI’s ingestion engine automatically applies Differential Privacy (ε, δ)-mechanisms, ensuring that individual PII (Personally Identifiable Information) cannot be reconstructed from the aggregate training set.
2. Regulatory Alignment & "The EU AI Act" Protocol
As a "Provider of High-Risk AI Training Data," Tekbod AI maintains a Technical Documentation Vault for every project to satisfy Article 11 of the EU AI Act:
Data Provenance & Lineage: We provide a 1:1 map of every data point's origin, the vetting level of the human annotator, and the specific "Alignment Taxonomy" used.
Bias Mitigation Reporting: We conduct automated and human-led "Statistically Significant Bias Audits" on all SFT (Supervised Fine-Tuning) datasets to ensure representative sampling across demographic protected groups.
3. Sovereign Bench Logic (Data Residency)
To satisfy the strictest interpretations of GDPR and CCPA, Tekbod AI offers Sovereign Data Sharding:
Local-to-Local Processing: If your data is generated in the EU/UK, it is labeled exclusively by our EU-Sovereign SME Bench on servers physically located within the EEA.
Cross-Border Transfer Restrictions: We utilize Standard Contractual Clauses (SCCs) enhanced by Supplementary Technical Measures (STMs) to prevent any "Schrems II" level legal vulnerabilities for our clients.
4. The "Right to be Forgotten" in Model Weights
One of the hardest challenges in AI is "Unlearning." Tekbod AI provides a Redaction-to-Retraining pipeline:
Hash-Linked Labels: If a data subject exercises their "Right to Erasure," Tekbod AI can instantly identify every prompt, ranking, or label influenced by that subject across your entire training corpus, allowing for precise data pruning and model recalibration.
COOKIE & TELEMETRY POLICY
• Privacy-First Analytics: We utilize server-side, anonymized telemetry for platform performance only. We employ zero third-party marketing pixels. Your interaction with the Tekbod AI portal is a closed-loop system.
Elite Cookie Policy: The "Zero-Tracker" Standard
• The "Zero-Tracker" Standard
Traditional B2B sites are cluttered with marketing pixels (Meta, LinkedIn, Google Ads). Tekbod AI removes these to ensure complete "Client Isolation."
Our Policy: Functional Sovereignty
Tekbod AI employs a Strictly Necessary cookie architecture. We do not use cookies for behavioral advertising, cross-site tracking, or third-party profiling.
✓ Category: Essential (Always On)
√Purpose: Session Authentication & Security.
√Function: Maintains your secure JWT (JSON Web Token) session and prevents Cross Site Request Forgery (CSRF) attacks.
√Retention: Deleted immediately upon logout or session expiry (max 4 hours).
✓ Category: Preference (Opt-in)
√Purpose: Interface Customization.
√Function: Remembers your Dark/Light mode settings and dashboard layout preferences.
✓ The "Anti-Pixel" Guarantee: "Tekbod AI does not use 'Shadow Profiles.' We have intentionally omitted tracking pixels from social media and ad networks to ensure that your presence on our platform remains invisible to the broader data-broker ecosystem."
Telemetry Policy: "Privacy-Preserving Analytics"
Policy: Metadata Minimization
Tekbod AI distinguishes between Platform Telemetry (how the app runs) and Data Telemetry (the content of your work). We have a Zero-Visibility policy on the latter.
A. Platform Health Telemetry
We collect anonymous, aggregated performance metrics to ensure 99.99% uptime.
• What we see: API latency, error rates, and regional load balancing data.
• What we don't see: Your IP address (it is masked/hashed at the edge) or your specific browser fingerprint.
B. Data Telemetry (The "Air-Gap" Promise)
• No Content Sampling: Unlike "freemium" tools, Tekbod AI never samples your prompts, images, or labels for our own model training.
• Differential Privacy: For project-level reporting, we utilize \epsilon-differential privacy (epsilon-differential privacy). This ensures that even if a summary report is generated, no individual data point or annotator's specific work can be reverse-engineered from the aggregate.
• End-to-End Auditability: All telemetry logs are available for Client inspection via our Transparency Portal, allowing your CISO to verify that no unauthorized data exfiltration has occurred.
ETHICAL DATA GOVERNANCE & CONTRIBUTOR PROTECTIONS
At Tekbod AI, we manage custom data collections by placing contributor dignity, explicit consent, and stringent technical governance at the heart of our operations. Our operational framework aligns directly with global standards—including the EU AI Act (Article 11), GDPR, CCPA, and ISO 27001/27701/42001 guidelines.
Below is how we operationalize contributor rights and ethical practices across custom collection lifecycles executed for enterprise clients:
1. Contributor Rights, Welfare & Fair Compensation
Fair Pay & Living Wage Guarantee: We adhere strictly to localized minimum wage and living wage standards. Pay rates reflect task complexity, specialized subject-matter expertise (e.g., engineering, legal, or domain-specific SMEs), and local cost-of-living metrics to prevent economic exploitation.
Informed, Multi-Layered Consent: Before joining any task, contributors undergo a clear consent process. They are presented with job-specific Non-Disclosure Agreements (NDAs) and transparent disclosures detailing exactly how their inputs will be processed, anonymized, and utilized for AI model training.
Right to Erasure & Account Control: Contributors retain total ownership of their personal footprint. If a contributor requests account deletion or revokes consent under GDPR/CCPA, our system executes automated account anonymization and triggers a Redaction-to-Retraining pipeline to prune hash-linked data points from project corpora.
2. Privacy Preservation & Zero-PII Collection Controls
Synthetic PII Masking Protocols: During custom collection tasks (e.g., voice, text, or prompt generation), contributors are instructed never to provide real Personally Identifiable Information (PII). When structural representation of sensitive data (like phone numbers or account structures) is required, contributors use randomized, synthetic formats.
Protocol-Level Anonymization: Data ingestion engines automatically apply -Differential Privacy mechanisms, ensuring individual PII cannot be reverse-engineered from aggregate client training sets.
Ephemeral Sandboxes: Contributor work occurs inside volatile memory sandboxes (Ephemeral Annotation Environments). Data is streamed, processed, and purged from local memory using cryptographic shredding standards (DoD 5220.22-M) upon task submission.
3. Ethical Sourcing, Diversity & Bias Mitigation
Demographic Representation & Inclusivity: We curate custom contributor panels targeting precise geographic, linguistic, and demographic distributions to address dataset bias at the source.
Statistically Significant Bias Audits: SFT (Supervised Fine-Tuning) and RLHF datasets undergo human-led and automated bias audits prior to client delivery to verify balanced sampling across protected demographic groups.
No Exploitative Labor Practices: We actively avoid reliance on unvetted micro-tasking pools. All contributors are vetted, onboarded, and continuously monitored for welfare standards and fair treatment.
4. Lineage, Data Provenance & Client Transparency
Full Data Auditability & Provenance: Every item delivered in a custom collection comes with a 1:1 Metadata Map detailing dataset lineage, vetting/qualification tiers of the contributor panel, and alignment taxonomies used.
Strict Regional Bench Sharding (Data Residency): To satisfy jurisdictional constraints (GDPR, EU AI Act), custom data collections can be routed exclusively through localized, regional contributor benches (e.g., EU-sovereign collections processed strictly within the EEA).
Air-Gap Guarantee: Custom data collected on behalf of a client remains completely isolated. Tekbod AI does not sample, cross-license, or repurpose client-commissioned datasets to train proprietary models.