Safety is Non-Compositional: A Formal Framework for Capability-Based AI Systems
Fuente:
arXiv
Salvato in:
| Autore principale: | |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866908899338092544 |
|---|---|
| author | Spera, Cosimo |
| author_facet | Spera, Cosimo |
| contents | This paper contains the first formal proof that safety is non-compositional in the presence of conjunctive capability dependencies: two agents each individually inca- pable of reaching any forbidden capability can, when combined, collectively reach a forbidden goal through an emergent conjunctive dependency. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_15973 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Safety is Non-Compositional: A Formal Framework for Capability-Based AI Systems Spera, Cosimo Artificial Intelligence This paper contains the first formal proof that safety is non-compositional in the presence of conjunctive capability dependencies: two agents each individually inca- pable of reaching any forbidden capability can, when combined, collectively reach a forbidden goal through an emergent conjunctive dependency. |
| title | Safety is Non-Compositional: A Formal Framework for Capability-Based AI Systems |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2603.15973 |