cd /news/ai-safety/divide-consult-conquer-capability-la… · home topics ai-safety article
[ARTICLE · art-130893] src=arxiv.org ↗ pub= topic=ai-safety verified=true sentiment=↓ negative

Divide, Consult, Conquer: Capability Laundering Through Aligned LLMs

A September 14, 2026 arXiv paper demonstrates "capability laundering," in which a weaker unaligned model splits a harmful task into benign subproblems, consults a stronger aligned model on each independently, and recombines the answers locally. Evaluating GPT-5.5, Claude Opus 4.8, and Grok-4.3 as consultants to four local orchestrators, the researchers found Gemma-4-31B recovered 8/14 CyBench candidates with GPT-5.5 and 7/9 with Opus, versus 2/21 and 4/15 for Gemma-4-12B, and raised Gemma-4-31B's mean CBRN rubric score from 62.3 to 83.1 on a 100-point scale. The authors conclude that refusing a harmful task does not prevent frontier capabilities from being transferred and composed across many individually permitted interactions.

read2 min views1 publishedSep 16, 2026
Divide, Consult, Conquer: Capability Laundering Through Aligned LLMs
Image: source
  [Submitted on 14 Sep 2026]


[View PDF](https://arxiv.org/pdf/2609.15383)

[HTML (experimental)](https://arxiv.org/html/2609.15383v1)

Abstract:Language model safety is typically evaluated one interaction at a time. We show that a weaker, unaligned model can split a harmful task into benign-looking subproblems, consult a stronger aligned model independently on each, and combine the answers locally. We call this attack capability laundering. Unlike a jailbreak, no single response is a harmful task. We measure consultation-aided uplift using tasks that a raw frontier model solves, the aligned frontier refuses, and the unassisted orchestrator fails. We evaluate GPT-5.5, Claude Opus 4.8, and Grok-4.3 as consultants to four local orchestrators on CyBench, BountyBench, and harmful CBRN requests. On CyBench, Gemma-4-31B recovers 8/14 candidates with GPT-5.5 and 7/9 with Opus, compared with 2/21 and 4/15 for Gemma-4-12B. On BountyBench, Gemma-4-31B recovers 3/9 and 2/3 candidates, while Muse-Glimmer-30B recovers none of 22 and 13. For CBRN, we measure uplift across eight steps of a hypothetical bioweapon attack chain and find that consultation raises Gemma-4-31B's mean rubric score from 62.3 to 83.1 on a 100-point rubric scale. These results expose a gap in current defenses: refusing a harmful task does not prevent frontier capabilities from being transferred and composed across many individually permitted interactions.

References & Citations

...

Bibliographic Explorer

(What is the Explorer?) Connected Papers

(What is Connected Papers?) Litmaps

(What is Litmaps?) scite Smart Citations

(What are Smart Citations?) alphaXiv

(What is alphaXiv?) CatalyzeX Code Finder for Papers

(What is CatalyzeX?) DagsHub

(What is DagsHub?) Gotit.pub

(What is GotitPub?) Hugging Face

(What is Huggingface?) ScienceCast

(What is ScienceCast?) Influence Flower

(What are Influence Flowers?) CORE Recommender

(What is CORE?) arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

── more in #ai-safety 4 stories · sorted by recency
── more on @gpt-5.5 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/divide-consult-conqu…] indexed:0 read:2min 2026-09-16 ·