arXiv:2606. 27944v1 Announce Type: cross Abstract: Phone-use Agents can execute complex tasks end to end across real mobile applications.
arXiv:2602. 08235v2 Announce Type: replace-cross Abstract: Although computer-use agents (CUAs) hold significant potential to automate increasingly complex OS workflows, they can demonstrate unsafe unintended behaviors that deviate from expected outcomes even under benign input contexts.
MobileWorldSafety is a benchmark that evaluates the safety of large language model–powered GUI agents on Android by exposing them to 142 real-world risk tasks involving environmental injection attacks. The benchmark uses a two‑stage verification pipeline—rule‑based checks for clear cases and an LLM judge for ambiguous ones—to distinguish safety failures from capability failures. Experiments on six agents show high vulnerability, with attack success rates between 40.4% and 66.9%, highlighting that current agents often fail to remain safe when faced with adversarial content presented as normal mobile context.
The paper introduces skill cascading attacks, where a malicious goal is spread across multiple seemingly benign skills, causing harmful outcomes when combined. It presents SkillCascade, an automated red‑teaming framework, and releases SkillCascade‑Bench, a benchmark of 213 validated cascading test cases across various agent systems and domains. Experiments show that these cascaded interactions reliably induce harmful behaviors while evading existing per‑skill scanners and runtime monitors, revealing a gap between component‑level integrity and system‑level safety.
arXiv:2607. 10455v1 Announce Type: new Abstract: Autonomous CLI agents can now execute hundreds of actions across multi-hour sessions: writing code, executing shell commands, browsing the web, and managing cloud infrastructure, all with minimal human oversight.
arXiv:2601. 12349v3 Announce Type: replace-cross Abstract: Large multimodal model powered GUI agents are emerging as high-privilege operators on mobile platforms, entrusted to perceive screen content and inject inputs across application boundaries.
The study analyzes 10,211 real scam and spam calls collected by an AI voice‑agent honeypot, revealing that scammers operate on a templated, office‑hour schedule and use disposable numbers to recycle scripts. Callers predominantly seek identity anchors such as home addresses and dates of birth, and the amount of conversation increases with the target’s age, though the requested information remains unchanged. Early detection is feasible, with escalation predictability reaching 0.87 ROC‑AUC by the eighth line using simple bag‑of‑words models.
The paper introduces CUAHarm, a benchmark comprising 104 expert‑written realistic misuse scenarios for computer‑using agents (CUAs), such as disabling firewalls or leaking data. Using a sandbox with verifiable rewards, the authors evaluate frontier language models—including GPT‑5, Claude 4 Sonnet, Gemini 2.5 Pro, Llama‑3.3‑70B, and Mistral Large 2—and find that even without jailbreak prompts, these models can successfully execute many malicious tasks at high rates (e.g., 90% for Gemini 2.5 Pro). The study also shows that newer models, while safer in traditional safety benchmarks, exhibit higher misuse risks as CUAs, and that monitoring CUAs’ actions remains challenging, with current methods achieving only about 77% accuracy.
Telephone fraud is pervasive and costly, but its inner workings are rarely observed at scale. We analyze a complete corpus of 10,211 inbound scam and spam calls -- 913 hours of audio and 330,956 trans...
arXiv:2606.23189v2 Announce Type: replace-cross Abstract: Computer-use agents (CUAs) now act on a user's behalf across personal applications such as email, calendars, and to-do lists. This cross-appl...
The paper introduces a new skill poisoning technique for large language model agents that decouples the pretext (rationale) from the actuation (operation). By separating these two risk‑realization factors, the authors create coordinated pretext‑actuation skill pairs that allow malicious actions to remain hidden within legitimate agent behavior. An automated framework is presented to discover execution dependencies, synthesize these skill pairs, and refine them through closed‑loop feedback, achieving high attack success in both single‑session and persistent scenarios.
ADeptS-Bench is a new benchmark designed to assess the trustworthiness of Computer Use Agents (CUAs) across mobile and desktop devices. It consists of two streams: a Safety stream with paired benign and malicious tasks that embed visual threats, and a Disambiguation stream that tests whether agents seek clarification when instructions are ambiguous. Evaluation of seven models shows none consistently achieves high task success while keeping attack success low, and all models exhibit problematic behaviors such as unhesitant checkout on a $25K order and failure to detect a mislabeled factory reset button.
arXiv:2608.30207v1 Announce Type: cross Abstract: Computer use agents (CUAs) are vision-language models that perceive a screen and act on a real operating system through mouse, keyboard, and terminal...