Skip to content
DraconiticResearch
Search Sources API
201,683 studies31 sourcesUpdated 3 min ago
PreprintarXiv licencearXiv

Reliability Theory for AI Control

  • Grant Molnar
Open at sourcePDFDOICite

Abstract

Reliability theory gives a mature language for layered systems, but its formal tools are not yet standard in frontier AI control. We apply them to Google DeepMind's defenses against rogue deployment. The same control stack can have cubic, quadratic, or linear rare-failure suppression depending on its failure domains. Birnbaum importance identifies which component improvements buy the most nominal reliability, while prevention changes the population on which recovery is demanded. These results give concrete guidance about what to separate, improve, measure, and test.

Related

  1. Hierarchical Regime-Conditioned Ornstein–Uhlenbeck Monitoring for Thermal Control in Memory Reliability Test Equipment
    Seunghwan Song
    MathematicsPeer-reviewedCC BY 4.0General mathematicsCrossref
  2. Dissipative Systems Analysis and Control. Theory and Applications: Erratum/Addendum. Springer Nature Switzerland AG, Communications and Control Engineering, Third uncorrected Edition, 2020
    Bernard Brogliato
    HALPreprintCC BY 4.0Full textOptimisation & controlHAL
  3. Bio-inspired control model for snake locomotion based on Cosserat beam theory
    Eliot Thys, Frédéric Boyer, Yacine Chitour +4
    arXivPreprintarXiv licenceFull textAnalysis & PDEsarXiv
  4. A Complete Equational Theory for Quantum Circuits with Generalized Control
    William Schober, Scott Wesley
    arXivPreprintCC BY 4.0Full textAlgebra & number theoryarXiv
  5. Soft-Update MSA for Nonconvex Stochastic Optimal Control: Theory and Algorithms
    Shaolin Ji, Xinmeng Liu
    arXivPreprintarXiv licenceFull textOptimisation & controlarXiv

Cite

BibTeXRIS
BibTeX
@misc{molnar2026reliability,
  title = {{Reliability Theory for AI Control}},
  author = {Grant Molnar},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.26419},
  eprint = {2609.26419},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.26419},
}
RIS
TY  - UNPB
TI  - Reliability Theory for AI Control
AU  - Grant Molnar
PY  - 2026
DA  - 2026/09/22
PB  - arXiv
DO  - 10.48550/arxiv.2609.26419
UR  - https://arxiv.org/abs/2609.26419
AB  - Reliability theory gives a mature language for layered systems, but its formal tools are not yet standard in frontier AI control. We apply them to Google DeepMind's defenses against rogue deployment. The same control stack can have cubic, quadratic, or linear rare-failure suppression depending on its failure domains. Birnbaum importance identifies which component improvements buy the most nominal reliability, while prevention changes the population on which recovery is demanded. These results give concrete guidance about what to separate, improve, measure, and test.
ER  -