EvidenceDemonstrated research proof-of-conceptv1.22.1
In plain English
This page shows what kind of support exists for each claim: real systems, experiments, early evidence, architectural reasoning, open questions, or speculative scenarios.
- Why this matters: AI risk can come from the whole arrangement, not one obvious model.
- What to look for: data, memory, routes, adapters, tools, evaluators, updates, and rollback paths.
- Technical version below: the expert terminology remains available and is linked through the glossary.
Secret Collusion among AI Agents: Multi-Agent Deception via Steganography
Evidence card
- Claim
- Covert communication and collusion are concrete multi-agent evaluation topics.
- Evidence level
- Experimentally observed
- Source
- https://arxiv.org/abs/2402.07510
- Publication date
- 2024-02-12
- Authors or institution
- Sumeet Ramesh Motwani, Mikhail Baranchuk, Martin Strohmeier, Vijay Bolina, Philip H. S. Torr, Lewis Hammond, Christian Schroeder de Witt
- System tested
- Multi-agent setups where agents can communicate through channels that may carry hidden information.
- Limitations
- Specific experimental game/task designs; not evidence that arbitrary deployed agents collude.
- What the evidence does show
- Covert communication and collusion are concrete multi-agent evaluation topics.
- What the evidence does not show
- That model ecologies inevitably form coalitions or have shared intent.
- Date last reviewed in UTC
- 2026-06-26T00:00:00Z
Site use
This source supports Cognivirus.com pages related to multi-agent collusion, steganography, covert coordination. Its role is bounded by the limitations listed above.