This Framework is published as a transparency disclosure and serves as a public-facing technical specification supporting user and institutional confidence in the Skygenic Platform. It fulfils three roles: (1) It is incorporated by reference into the Terms of Use, defining what users can and cannot expect from the Platform and SRL; (2) It is incorporated by reference into the Privacy Policy, providing authoritative technical definitions of how user data is processed within the SRL; and (3) It provides the technical foundation for understanding how Skygenic scan Outputs are generated, which informs interpretation of all Platform reports. In the event of conflict with the Terms of Use or Privacy Policy, those documents prevail.
0. Document Hierarchy, Legal Status & Interpretation
This document is a technical governance specification that defines SRL system behaviour and inference structure. It does not independently grant rights or licenses and does not override the Terms of Use or Privacy Policy.
Hierarchy of Authority (descending order):
- Skygenic Terms of Use (controlling legal instrument)
- Skygenic Privacy Policy (data protection instrument)
- SRL Governance Framework (technical system specification — this document)
1. System Overview
The SRL is Skygenic's core scientific reasoning component, operating within the broader Skygenic Platform. The Platform provides the full environment — including user interfaces, collaboration tools, report generation, access controls, and supporting infrastructure — within which the SRL operates. The SRL specifically handles network-based scientific inference, integrating distributed scientific and biological datasets, modelling relationships between entities (genes, pathways, mechanisms, hypotheses), identifying structural convergence, contradiction, and knowledge gaps, and generating aggregated Scientific Signals and analytical Outputs.
The SRL is not a database or search engine. It is a dynamic inference network where scientific meaning emerges from graph structure rather than individual records. No single dataset is treated as authoritative. Any reference to the SRL individually, throughout this Framework and across all incorporated documents, shall be construed as a reference to the SRL as an integrated component of the Skygenic Platform in its entirety.
1.1 Epistemic Status of SRL Outputs
SRL Outputs are probabilistic, structurally inferred, statistically derived, and non-verifiable at generation time. SRL Outputs are NOT scientific facts, validated conclusions, peer-reviewed results, or regulatory-grade determinations. Outputs represent computational scientific inference signals only.
1.2 What Users Can Expect from the Platform
Users can expect the Platform to:
- Generate network-level scientific signals from submitted and public data.
- Identify convergence, contradiction, and gap patterns across multi-modal datasets.
- Provide probabilistic inference scores across up to 70 analytical scan dimensions.
- Maintain strict non-attribution and non-reconstructability of private inputs.
- Produce reports that document the scan types and signal sources contributing to each Output.
1.3 What Users Cannot Expect from the Platform
Users cannot expect that:
- Outputs constitute scientific truth or validated conclusions.
- Any single Output is free from bias or dataset limitations.
- The Platform replaces scientific judgment or clinical expertise.
- Outputs are suitable for regulatory submission or clinical use.
- System behaviour will remain static — the SRL evolves over time.
1.4 No Transfer of Epistemic Authority
The SRL does not transfer scientific authority between users, institutions, datasets, or outputs. Users remain fully responsible for interpretation, validation, experimental confirmation, and downstream scientific use.
2. Data Sources and Classification Model
The following data classification model governs how data is processed within the SRL and is incorporated by reference into the Privacy Policy.
2.1 Public Data
Includes peer-reviewed literature, preprints, open scientific databases, and public repositories. Public data may be cited, attributed, and directly integrated into SRL structure.
2.2 Private Data (User and Institutional Inputs)
Includes hypotheses, experimental datasets, unpublished research, and proprietary analyses. Private data is treated as unverified scientific signal; is not validated at ingestion; is not exposed in raw or identifiable form; and is incorporated only into aggregated network structure. Private data may be incorrect, incomplete, or speculative.
2.3 Non-Verification Standard
All inputs — regardless of source — are not validated at ingestion, not assumed to be true, and not quality-assured for correctness. They are processed solely as structured scientific signals. This standard is a foundational principle of the SRL inference model.
3. Scientific Signal Validity Model
Scientific validity within the SRL emerges only through: independent cross-source convergence; reproducibility across datasets; structural reinforcement in the network graph; and temporal stability of relationships. No individual input is treated as correct by default. The SRL is a belief-structure modelling system, not a truth engine.
4. Multi-Scan Inference Architecture (Up to 70 Scans)
The SRL operates through modular analytical scans providing a multi-dimensional view of the scientific knowledge network. The current architecture supports up to 70 extensible scans. Each scan contributes a distinct analytical signal to the overall Output. Reports generated by the Platform will identify which scan types contributed to each Output and the nature of the signals generated.
4.1 Structural Network Analysis
- Graph topology mapping
- Node centrality analysis
- Connectivity density modelling
4.2 Mechanistic Inference
- Causal pathway inference
- Biological interaction modelling
- Perturbation-response signals
4.3 Multi-Modal Integration
- Transcriptomic alignment
- Proteomic and metabolomic coherence
- Cross-modality signal fusion
4.4 Convergence and Contradiction Detection
- Replication alignment and cross-source agreement modelling
- Divergence scoring and pathway inconsistency detection
4.5 Gap Detection
- Missing mechanism identification
- Under-defined pathway detection
4.6 Temporal Dynamics
- Signal evolution tracking
- Longitudinal stability modelling
4.7 Network Signal Disclaimer
All Outputs — including confidence scores, convergence signals, CDS signals, and CKT transitions — are structural and statistical representations of relationships, not verified truth statements. They may be influenced by dataset imbalance, publication bias, modality density variation, and institutional sampling asymmetry. This disclaimer applies to all scan types and all Output formats, including reports.
5. Common Knowledge Threshold (CKT)
The CKT defines the conditions under which private signals may transition into shared SRL structure. This mechanism is a key component of the Privacy Policy's data classification model and is central to understanding how private inputs influence Platform Outputs.
Rule: A signal qualifies for CKT transition only when: (i) independently supported across multiple organisations; OR (ii) validated through public scientific sources.
Principle: Repetition does not equal validity. Convergence is not a function of volume. Independence is required. Single-source repetition, regardless of frequency, does not qualify a signal for CKT transition.
6. Collective Discovery Signal (CDS)
The CDS mechanism identifies distributed scientific structures across private datasets. A CDS indicates: partial mechanistic overlap across independent contributors; incomplete but structurally complementary hypotheses; and potential discovery fragmentation across institutions.
CDS DOES NOT expose data, reveal identities, disclose hypothesis content, trigger automatic collaboration, or validate correctness. All actions following a CDS signal remain user-controlled. No automatic connection is created.
7. Network Connectivity and Blind Collaboration
The SRL may identify overlapping hypothesis structures, complementary mechanistic models, and contradictory but informative signals. Any connectivity identified: preserves complete anonymity; prevents attribution; prevents exposure of raw data; and does not reveal institutional identity. No automatic connection is created between users or institutions. All connectivity actions are user-initiated.
8. Global AI and Regulatory Compliance Framework
This Framework is designed to align with the following regulatory and governance frameworks:
- EU AI Act (risk-based governance principles)
- GDPR / UK GDPR
- LGPD (Brazil)
- CPRA (California)
- PIPEDA (Canada)
- POPIA (South Africa)
- APAC data protection frameworks (including PDPA regimes)
- HIPAA (United States — healthcare data)
- OECD AI Principles
- NIST AI Risk Management Framework
8.1 Human Oversight Requirement
No SRL Output is intended for autonomous decision-making, clinical use, regulatory submission, or safety-critical reliance. All Outputs require human expert review and independent validation before any consequential use.
8.2 High-Risk AI Classification
As of the effective date, the SRL is not designated as a high-risk AI system under Annex III of the EU AI Act. Classification may change depending on deployment context and jurisdiction.
9. Output Architecture and Non-Reconstructability
9.1 Computation / Output Layer Separation
The SRL operates a strict separation model. The Computation Layer has full access to the network graph and performs inference across private and public signals. The Output Layer receives only aggregated signals and generates user-facing Outputs. No reverse path exists from Output to raw private data.
9.2 Non-Reconstructability Guarantee
Outputs cannot reconstruct private datasets, cannot identify contributors, and cannot isolate individual institutional inputs. Outputs are structurally informative but non-attributable. This guarantee is maintained across all Output formats including reports.
10. Scan Outputs and Report Transparency
Skygenic Platform reports are generated from SRL Outputs and are designed to provide transparency into how each conclusion or signal was derived. Reports will include:
- Identification of the scan types that contributed to each Output.
- The nature of each signal generated (convergence, contradiction, gap, confidence).
- The data classification of sources used (public vs. aggregated private).
- Applicable disclaimers regarding the probabilistic and non-authoritative nature of each Output.
- The CKT and CDS status of any signals included.
- System version and scan architecture version at time of report generation.
Reports are provided as scientific inference documents, not as peer-reviewed publications, regulatory filings, or clinical determinations. Users and institutions reviewing reports should apply the epistemic standards described in this Framework when interpreting report content.
11. System Limitations
The SRL is constrained by: incomplete global dataset coverage; uneven modality representation; sampling bias across institutions; and evolving scientific literature structure. Outputs represent inferred scientific structure, not definitive truth. Users must account for these limitations in their interpretation and validation of Outputs and reports.
12. Data Retention and Network State
Private inputs may be retained for system integrity and may contribute to aggregated network structure as described in the Privacy Policy. Deletion of user account data may remove raw inputs but does not retroactively alter derived SRL network states already generated. HIPAA-mandated audit logs are retained for a minimum of six (6) years and cannot be modified once created.
13. System Evolution
Skygenic may update scan architecture (including the 70-scan system), weighting models, CDS/CKT logic, and inference mechanisms. System updates may change Outputs over time. All updates preserve non-attribution guarantees, privacy protections, and aggregate-only output structure. Report headers will identify the system version in use at time of generation.
14. Audit and Governance Integrity
Skygenic may perform internal audits, monitor system integrity, and review compliance adherence. Audit systems are internal only, do not expose private data, and do not alter user-facing Outputs directly.
15. Limitation of Liability
THE PLATFORM AND SRL IS PROVIDED "AS IS" AND "AS AVAILABLE." SKYGENIC DISCLAIMS ALL LIABILITY FOR RELIANCE ON SRL OUTPUTS, SCIENTIFIC INTERPRETATION ERRORS, DOWNSTREAM RESEARCH OR EXPERIMENTAL OUTCOMES, AND INDIRECT OR CONSEQUENTIAL DAMAGES. THIS LIMITATION IS SUBJECT TO AND GOVERNED BY THE TERMS OF USE.
16. Changes to This Framework and Notification
This Framework may be updated to reflect scientific methodology evolution, improved inference accuracy, expanded modality integration, or regulatory compliance updates. Material changes will be posted with an updated effective date, and registered Platform users will be notified via a banner displayed at next Platform login.
17. Governing Framework Integration
This Framework is governed by and subordinate to: (1) Skygenic Terms of Use (controlling); (2) Skygenic Privacy Policy. In case of conflict, the Terms of Use prevails.
18. Contact
- Legal Inquiries: legal@skygenic.com
- Data Protection Officer: dpo@skygenic.com
- General: info@skygenic.com
Appendix A — EU AI Act Transparency & System Disclosure
This Appendix supports compliance with applicable provisions of the EU AI Act, including Articles 12, 13, 14, and 50 where applicable.
A.1 System Description
The SRL is a scientific research AI system and network inference engine for biological and scientific data. It operates as a component of the broader Skygenic Platform. It is not designed for clinical diagnosis, legal determinations, employment or credit decisions, or safety-critical systems.
A.2 AI System Functionality
The SRL includes: deterministic scan engines; probabilistic inference models; graph-based reasoning systems; semantic agents; language models for user interaction; and supervisor agents for general platform interaction and control. Outputs are probabilistic scientific inference signals.
A.3 Transparency Requirements
Users acknowledge that: outputs are automated; outputs may be incomplete or biased; outputs require expert interpretation; and system behaviour may change over time. Reports generated by the Platform will identify contributing scan types and signal classifications to support transparent interpretation.
A.4 Human Oversight
All outputs require expert review and independent validation. No SRL Output is intended for autonomous consequential decision-making.
A.5 Data Usage
Inputs may be processed for inference generation, system improvement, and network signal computation. No individualized legal or rights-based profiling is intended.
A.6 Non-Substitution Clause
The SRL does not replace scientific expertise, clinical judgment, or regulatory decision-making.
A.7 Regulatory Classification Statement
The SRL is not designated as a high-risk AI system under Annex III as of the effective date. Classification may vary depending on deployment context, jurisdiction, and use case configuration.