Speech-driven large language models (LLMs) are increasingly accessed through speech interfaces, introducing new security risks via open acoustic channels. We present Sirens' Whisper (SWhisper), the first practical framework for covert prompt-based attacks against speech-driven LLMs under realistic black-box conditions using commodity hardware. SWhisper enables robust, inaudible delivery of arbitrary target baseband audio-including long and structured prompts-on commodity devices by encoding it into near-ultrasound waveforms that demodulate faithfully after acoustic transmission and microphone nonlinearity. This is achieved through a simple yet effective approach to modeling nonlinear channel characteristics across devices and environments, combined with lightweight channel-inversion pre-compensation. Building on this high-fidelity covert channel, we design a voice-aware jailbreak generation method that ensures intelligibility, brevity, and transferability under speech-driven interfaces. Experiments across both commercial and open-source speech-driven LLMs demonstrate strong black-box effectiveness. On commercial models, SWhisper achieves up to 0.94 non-refusal (NR) and 0.925 specific-convincing (SC). A controlled user study further shows that the injected jailbreak audio is perceptually indistinguishable from background-only playback for human listeners. Although jailbreaks serve as a case study, the underlying covert acoustic channel enables a broader class of high-fidelity prompt-injection and commandexecution attacks.

The rapid adoption of large language models (LLMs) in enterprise systems exposes vulnerabilities to prompt injection attacks, strategic deception, and biased outputs, threatening security, trust, and fairness. Extending our adversarial activation patching framework (arXiv:2507.09406), which induced deception in toy networks at a 23.9% rate, we introduce the Unified Threat Detection and Mitigation Framework (UTDMF), a scalable, real-time pipeline for enterprise-grade models like Llama-3.1 (405B), GPT-4o, and Claude-3.5. Through 700+ experiments per model, UTDMF achieves: (1) 92% detection accuracy for prompt injection (e.g., jailbreaking); (2) 65% reduction in deceptive outputs via enhanced patching; and (3) 78% improvement in fairness metrics (e.g., demographic bias). Novel contributions include a generalized patching algorithm for multi-threat detection, three groundbreaking hypotheses on threat interactions (e.g., threat chaining in enterprise workflows), and a deployment-ready toolkit with APIs for enterprise integration.

State-of-the-art (SOTA) fact-checking systems combat misinformation by employing autonomous LLM-based agents to decompose complex claims into smaller sub-claims, verify each sub-claim individually, and aggregate the partial results to produce verdicts with justifications (explanations for the verdicts). The security of these systems is crucial, as compromised fact-checkers can amplify misinformation, but remains largely underexplored. To bridge this gap, this work introduces a novel threat model against such fact-checking systems and presents \textsc{Fact2Fiction}, the first poisoning attack framework targeting SOTA agentic fact-checking systems. Fact2Fiction employs LLMs to mimic the decomposition strategy and exploit system-generated justifications to craft tailored malicious evidences that compromise sub-claim verification. Extensive experiments demonstrate that Fact2Fiction achieves 8.9\%--21.2\% higher attack success rates than SOTA attacks across various poisoning budgets and exposes security weaknesses in existing fact-checking systems, highlighting the need for defensive countermeasures.