LLM Side-Channel Attack Exposes User Conversations
Article Content
Browse articles 
Microsoft researchers reported that a side-channel attack can enable unauthorized parties to infer the topics being discussed with large language models (LLMs). Models from providers such as Anthropic, AWS, DeepSeek, and Google remain vulnerable, affecting both personal users and enterprise communications. The attack exploits indirect signals like power consumption and timing to extract sensitive information.
Ask AI about this cluster
Answers cite the sources they use
Updated 225d ago How this analysis works
More articles in this cluster (2)
Following this threat?
Track AWS in your own feed — you're alerted when they show up in new reporting, leak sites or exploitation.
Free account · no card needed
Continue Reading
New Spectre v2 Variant BTR Exposes JIT Engines to Data Leaks Researchers from Vrije Universiteit Amsterdam and Scuola Superiore Sant’Anna have disclosed a new Spectre v2 variant called Branch Target Reuse (BTR), affecting JIT compilers in browsers and operating system kernels across Intel, AMD, and Arm CPUs. This attack exploits stale indirect branch prediction entries… 
Critical BeyondTrust 9.3 Vulnerability Actively Exploited by Threat Actor A critical vulnerability (CVE-2026-102764) in BeyondTrust Remote Support has been discovered and is actively being exploited by the threat group ARMORY CASTLE, reportedly linked to China. The vulnerability, which has a CVSS score of 9.3, allows unauthenticated attackers to upload files and potentially compromise host…