Metraipolis › archive › dc_af498055ef4b

Capability-Gated Language Models: Security Composes, Utility Does Not — arXiv:2609.00445

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.00445
License (as archived)
CC0-1.0
Archived
2026-10-17T06:33:22.034Z
SHA-256 of the archived text
115619b3936ce689f130603c128534fab3bb0df16f31405da09093651d901b56
Stable URL of this page
https://metraipolis.com/documento/dc_af498055ef4b

How to cite

Capability-Gated Language Models: Security Composes, Utility Does Not — arXiv:2609.00445 — arXiv (en), https://arxiv.org/abs/2609.00445 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_af498055ef4b (Metraipolis archive dc_af498055ef4b, sha256 115619b3936ce689f130603c128534fab3bb0df16f31405da09093651d901b56)
BibTeX
@misc{dcaf498055ef4b,
  title = {Capability-Gated Language Models: Security Composes, Utility Does Not — arXiv:2609.00445},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.00445},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_af498055ef4b; sha256 115619b3936ce689f130603c128534fab3bb0df16f31405da09093651d901b56},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.00445. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Deployed language model safeguards (safety fine-tuning, filtering, unlearning) vary by principal only outside the model weights: filters are reconfigured, tiers are multiplied, and artefacts are reissued; inside one set of weights every request meets the same model configuration. This motivates us to define capability-gated deployment: per-principal access control inside one set of weights, whose configurations form a lattice - meets accumulate a principal's restrictions and joins pool a coalition's reach. We instantiate it by sparse rank gating over an existing nested-factorisation mechanism, guide profile search with one-pass attribution, and read every result once from a pre-registered held-out split. Security approximately composes: provably exactly at meets under a monotone-elicitation assumption we falsify pointwise. In two lineages the median held-out meet deepens suppression; the one effect surviving correction strengthens it. Utility does not: individually harmless profiles can compose to retention and fluency damage, and no compositional bound exists.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-02 in cs.AI).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.00445 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T06:15:12.780Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).