New research on AI and security, selected for relevance to language models, AI agents and cybersecurity. Summaries are the authors’ or publisher’s own words, linked to the original.
Peer-reviewed
Supervised deep learning for efficient cryo-EM image alignment in drug discovery with cryoPARES
Programmable functional RNA design with a virtual scientist
AI is ditching hallucinated references — and other bad habits
Estimating malaria burden in sub-Saharan Africa under intervention disruption using satellite-derived representations
Hypothalamic volume and cell-type changes in ageing and Alzheimer’s disease
AI must be held accountable — value statements won’t protect human dignity
New preprints (not yet peer-reviewed)
PIPETTE: Prompt Injection Purple-Teaming for Robust LLM Agents
Large language model (LLM) agents routinely interact with untrusted external data, exposing them to indirect prompt injection (IPI) attacks that hijack agents into executing attacker-chosen tasks. To uncover these vulnerabilities at scale, automated red-teaming methods often leverage reinforcement learning (RL) to generate attacks that achieve high attack success rates (ASR).
Colosseum: Auditing Collusion in Cooperative Multi-Agent Systems
Multi-agent systems, where LLM agents communicate through free-form language, enable sophisticated coordination for solving complex cooperative tasks. This surfaces a unique safety problem when a group of agents forms a coalition and colludes to pursue secondary goals and degrade the joint objective.
ACaM: Automated Capability Modeling for Agentic AI Security
In agentic AI systems, an LLM invokes tools that hold standing access to credentials, private data, filesystems, and external services. Such agents are increasingly deployed without an answer to a basic question: what can an agent access, exfiltrate, and propagate, independent of what it is designed for or is observed to do?
SpecLook: Speculating Where to Look for Fast and Reliable Agentic MLLMs
Agentic multimodal large language models (MLLMs) that think with images interleave reasoning with zoom-in tool calls. However, this sequential loop is slow, and its answers carry no reliability score.
LLM-Guided Cyber-Physical Attack Path Identification in Smart Grids
Identifying high-impact attack paths in cyber-physical power systems is critical for improving grid resilience. Existing methods typically rely on either structural topology analysis or simulation-based cascading-failure models.
LLM-Guided Cyber-Physical Attack Path Identification in Smart Grids
Identifying high-impact attack paths in cyber-physical power systems is critical for improving grid resilience. Existing methods typically rely on either structural topology analysis or simulation-based cascading-failure models.
AutoPI: Taxonomy-Guided Adaptive Red Teaming for Prompt Injection in LLM Agents
Prompt injection is the dominant unsolved security risk for LLM agents, and the community now agrees that robustness must be measured against adaptive attackers rather than fixed benchmark injections. Many adaptive red teamers exist, and most are built to maximize attack success rate (ASR).
AutoPI: Taxonomy-Guided Adaptive Red Teaming for Prompt Injection in LLM Agents
Prompt injection is the dominant unsolved security risk for LLM agents, and the community now agrees that robustness must be measured against adaptive attackers rather than fixed benchmark injections. Many adaptive red teamers exist, and most are built to maximize attack success rate (ASR).
