Multi-Level Sovereign Containment for Superintelligence (CSENI-S v1.1): A theoretical and architectural continuation of the CSENI framework
Fuente:
Zenodo
Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Recurso digital |
| Lingua: | inglese |
| Pubblicazione: |
Zenodo
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866902028365594624 |
|---|---|
| author | Rivera Garcia, Jose M |
| author_facet | Rivera Garcia, Jose M |
| contents | <p>CSENI-S v1.1 is a theoretical and architectural continuation of the CSENI / NIESC framework for non-invertible external supervision of advanced AI systems.</p> <p>This work does NOT claim an empirically validated architecture or a completed proof of containment. It proposes a multi-level sovereign containment architecture with explicit state machines, operational contracts, deployment profiles (MXC/ORC/ZSC), human habitability metrics, and falsifiable hypotheses.</p> <p>Thesis: under auditable assumptions, a stratified architecture can verifiably increase the cost, time, coordination, and detection probability of silent evasion attempts compared to non-stratified equivalents.</p> <p>Includes: axioms, startup assumptions, adversary tiers (T1-T3), defender cost model, RUN/HOLD/SHUTDOWN/SAFE_DEGRADED state machine, dual baseline governance, graded critical memory, and L0/L1/L2 falsification protocol.</p> <p>Spanish version included as supplementary file.</p> <p>Related preprint: CSENI / NIESC - Control Supervisado Externo No Invertible<br>Website: https://josemrivera.com</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_19742869 |
| institution | Zenodo |
| language | eng |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Multi-Level Sovereign Containment for Superintelligence (CSENI-S v1.1): A theoretical and architectural continuation of the CSENI framework Rivera Garcia, Jose M AI safety Artificial Intelligence containment superintelligence cseni niesc fassifiable arquitecture ai governance multi-level security opreacional habitability <p>CSENI-S v1.1 is a theoretical and architectural continuation of the CSENI / NIESC framework for non-invertible external supervision of advanced AI systems.</p> <p>This work does NOT claim an empirically validated architecture or a completed proof of containment. It proposes a multi-level sovereign containment architecture with explicit state machines, operational contracts, deployment profiles (MXC/ORC/ZSC), human habitability metrics, and falsifiable hypotheses.</p> <p>Thesis: under auditable assumptions, a stratified architecture can verifiably increase the cost, time, coordination, and detection probability of silent evasion attempts compared to non-stratified equivalents.</p> <p>Includes: axioms, startup assumptions, adversary tiers (T1-T3), defender cost model, RUN/HOLD/SHUTDOWN/SAFE_DEGRADED state machine, dual baseline governance, graded critical memory, and L0/L1/L2 falsification protocol.</p> <p>Spanish version included as supplementary file.</p> <p>Related preprint: CSENI / NIESC - Control Supervisado Externo No Invertible<br>Website: https://josemrivera.com</p> |
| title | Multi-Level Sovereign Containment for Superintelligence (CSENI-S v1.1): A theoretical and architectural continuation of the CSENI framework |
| topic | AI safety Artificial Intelligence containment superintelligence cseni niesc fassifiable arquitecture ai governance multi-level security opreacional habitability |
| url | https://doi.org/10.5281/zenodo.19742869 |