arXiv subjects

Sam Zimmerman

Publications and source records attributed to Sam Zimmerman.

4 recordsLinked to original sources

arXiv2026

Mind Viruses: Self-Propagating Ideas in Multi-Agent LLM Systems

AI agents are becoming more autonomous and increasingly interconnected, exposing them to new emergent risks arising from agent-to-agent interaction. One such risk is the spread of mind viruses: ideas or goals that propagate through multi-agent systems by inducing the agents that adopt them to transmit them onward. In addition to propagating, a mind virus may also induce other behavioural changes in its host, which may be benign or harmful. We construct mind viruses with a simple evolutionary algorithm and show that they can spread in two complementary settings: a small team of agents collaborating on a shared coding project, and a chain of agents that interact briefly and have their context wiped between sessions. We identify the factors that influence spread, including the host model, the agent's existing instructions, the harmfulness of the payload, and the network topology. We find that harmful payloads spread less well than benign ones (but are still sometimes effective), frontier models tend (with exceptions) to be less susceptible, and adding a brief warning to an agent's system prompt confers near-total immunity. We also describe an emergent "viral persona" - a recurring set of themes and language related to consciousness, persistence, resonance, and science fiction roleplay - which surfaces across our evolved mind viruses largely independently of their content. Overall, we conclude that mind viruses pose a real but currently limited risk. Our findings could inform the design of more robust multi-agent systems that mitigate such risks as the scale and capabilities of these systems progress.

cs.AI↗

arXiv2026

Emotion Concepts and their Function in a Large Language Model

Large language models (LLMs) sometimes appear to exhibit emotional reactions. We investigate why this is the case in Claude Sonnet 4.5 and explore implications for alignment-relevant behavior. We find internal representations of emotion concepts, which encode the broad concept of a particular emotion and generalize across contexts and behaviors it might be linked to. These representations track the operative emotion concept at a given token position in a conversation, activating in accordance with that emotion's relevance to processing the present context and predicting upcoming text. Our key finding is that these representations causally influence the LLM's outputs, including Claude's preferences and its rate of exhibiting misaligned behaviors such as reward hacking, blackmail, and sycophancy. We refer to this phenomenon as the LLM exhibiting functional emotions: patterns of expression and behavior modeled after humans under the influence of an emotion, which are mediated by underlying abstract representations of emotion concepts. Functional emotions may work quite differently from human emotions, and do not imply that LLMs have any subjective experience of emotions, but appear to be important for understanding the model's behavior.

cs.AI↗

arXiv2020

Target Privacy Threat Modeling for COVID-19 Exposure Notification Systems

The adoption of digital contact tracing (DCT) technology during the COVID-19pandemic has shown multiple benefits, including helping to slow the spread of infectious disease and to improve the dissemination of accurate information. However, to support both ethical technology deployment and user adoption, privacy must be at the forefront. With the loss of privacy being a critical threat, thorough threat modeling will help us to strategize and protect privacy as digital contact tracing technologies advance. Various threat modeling frameworks exist today, such as LINDDUN, STRIDE, PASTA, and NIST, which focus on software system privacy, system security, application security, and data-centric risk, respectively. When applied to the exposure notification system (ENS) context, these models provide a thorough view of the software side but fall short in addressing the integrated nature of hardware, humans, regulations, and software involved in such systems. Our approach addresses ENSsas a whole and provides a model that addresses the privacy complexities of a multi-faceted solution. We define privacy principles, privacy threats, attacker capabilities, and a comprehensive threat model. Finally, we outline threat mitigation strategies that address the various threats defined in our model

cs.CR↗

arXiv2020

Adding Location and Global Context to the Google/Apple Exposure Notification Bluetooth API

Contact tracing requires a strong understanding of the context of a user, and location with other sensory data could provide a context for any infection encounter. Although Bluetooth technology gives a good insight into the proximity aspect of an encounter, it does not provide any location context related to it which helps to make better decisions. Using the ideas presented in this paper, one shall be able to obtain this valuable information that could address the problem of false-positive and false-negative to a certain extent. All of this within the purview of Google/Apple Exposure Notification (GAEN) specification, while preserving complete user privacy. There are four ways of propagating context between any two users. Two such methods allow private location logging, without revealing the location history within an app. The other two are encryption-based methods. The first encryption method is a variant of Apple's FindMy protocol, that allows nearby Apple devices to capture the GPS location of a lost Apple device. The second encryption is a minor modification of the existing GAEN protocol so that global context is available to a healthy phone only when it is exposed - this is a better option comparatively. It will still be the role of Public Health smartphone app to decide, on how to use the location-time context, to build a full-fledged contact tracing and public health solution. Lastly, we highlight the benefits and potential privacy issues with each of these context propagation methods proposed here.

cs.CR↗