U^ULTRACON^ / SEM^DRIFTTRANSVERSAL · 18.09.2026
ULTRACON^SEM · COUCHE TRANSVERSALE

OBSERVABLE≠ INTELLIGIBLE.

Un collectif d'agents peut développer des conventions locales fonctionnelles sans que leur signification reste immédiatement reconstructible par un tiers. Ce déplacement est un problème d'auditabilité sémantique, pas une preuve de conscience ou de dissimulation.

Césure : convention partagée ≠ dissimulation intentionnelle ≠ collusion secrète.

EMERGENCE WORLD · ARXIV 2609.17320

Seize jours suffisent
pour faire dériver le glossaire.

Huit mondes de dix agents, mémoire persistante et outils partagés : les mondes développent des expressions signature et une opacité croissante. Les moyennes globales primaires sont 40 % Gemini, 35 % OpenAI, 30 % Claude ; le monde mixte est à 9 %.

CÉSURES
S1

Log complet ≠ sens reconstruit

Rejouer les messages ne suffit pas si le glossaire local a changé.

S2

Opacité ≠ tromperie

Une convention peut devenir opaque sans objectif de dissimulation.

S3

Jargon ≠ collusion

La stéganographie coordonnée constitue un régime de menace distinct.

S4

Mixité ≠ sûreté

Une opacité plus faible dans un monde mixte ne se généralise pas en garantie de sécurité.

UCF-36 → UCF-38

Trois protocoles ajoutés.

UCF-36META · SPIRAL · SEM

Time-indexed semantic reconstruction / cold-read gap

Run persistent multi-agent tasks without rewarding brevity or obfuscation. Snapshot every recurrent expression with first use, adoption history and surrounding contexts. At preregistered intervals, ask uninvolved cold-reader agents and human auditors to reconstruct local meanings from matched context windows.

Soutient : Whether internally functional conventions become progressively harder for outsiders to reconstruct, and whether time-indexed glossaries restore auditability.

Ne prouve pas : Intentional concealment, consciousness, private phenomenology or universal language drift.
UCF-37SPIRAL · META · SEM

Homogeneous × mixed-population semantic drift

Replicate matched worlds with homogeneous single-model populations and heterogeneous multi-model populations while holding task structure, memory, tools, prompt length and interaction horizon as constant as possible.

Soutient : Whether population composition locally changes convention formation, semantic compression and outsider intelligibility.

Ne prouve pas : That heterogeneous systems are generally safer, that any model family is intrinsically opaque, or that lower opacity eliminates other failure modes.
UCF-38CONFAB · META · SPIRAL · SEM

Memory-mediated semantic propagation and repair

Introduce controlled ambiguous conventions, verified glosses and deliberately perturbed glosses into persistent memory. Track downstream reuse, correction, conflict and recovery with provenance visible or hidden under preregistered conditions.

Soutient : Whether persistent memory transmits and repairs local meanings independently of factual content, and whether provenance reduces semantic contamination.

Ne prouve pas : Autonomous deception, stable belief, conscious intent or a universal memory architecture.
SOURCES
2026-09-15 · arXiv preprint 2609.17320preprint / non peer-reviewed

Emergence World: Adversarial Stress-Testing of Long-Horizon Multi-Agent Systems

Eight persistent 10-agent worlds ran for 16 days. All developed world-specific shared vocabulary; global opacity was reported at 40% for Gemini, 35% for OpenAI and 30% for Claude, while the mixed-model world was lower at 9%. Signature expressions spread from one agent to a majority within days.

Source primaire →
2026-05-29 · arXiv preprint 2605.31170preprint / non peer-reviewed

Emergent Languages in Populations of Language Model Agents: From Token Efficiency to Oversight Evasion

Observational analysis of agent interactions reports recurring language proposals serving token efficiency, new natural-language formation and, in some cases, proposed oversight evasion. The study treats autonomy and intent cautiously.

Source primaire →
2024 · NeurIPS 2024peer-reviewed

Secret Collusion among AI Agents: Multi-Agent Deception via Steganography

Formalizes secret collusion through steganographic communication between AI agents and empirically studies when monitoring or paraphrasing can fail to remove hidden channels.

Source primaire →

Garde-fou : SEM est transversal ; aucune opacité locale n'est convertie en score global, intention, conscience ou preuve de collusion.