Attackers who probe large language models rarely give up after one refusal. They reframe, build context across turns, adopt personas, and escalate gradually. New research from Cisco’s AI threat intelligence team finds that the safety benchmarks used across the industry miss almost all of this behavior, and the gap between published scores and observed resilience…
Tag: probe
AI, Cybersecurity, Global Security News, privacy
Agent’s claims on WhatsApp access spark security concerns
A US agent claimed WhatsApp encryption is fake and Meta can access messages; the probe was abruptly shut, raising security concerns. A US agent claimed WhatsApp encryption is fake, alleging Meta accesses all unencrypted messages, but Commerce Department abruptly shut the probe, leaving leaders questioning if consumer apps are safe for sensitive business decisions. In…
AI, Data Breaches, Global Security News
Scanning for AI Models, (Tue, Apr 14th)
Starting March 10, 2026, my DShield sensor started getting probe for various AI models such as claude, openclaw, huggingface, etc. Reviewing the data already reported by other DShield sensors to ISC, the DShield database shows reporting of these probes started that day and has been active ever since. Based on what we currently have reported,…
Global Security News, malware
Fraud Investigation Reveals Sophisticated Python Malware
Sophisticated Python malware uncovered in fraud probe shows obfuscation, disposable infrastructure
AI, Compliance, Data Breaches, Europe, Global Security News, privacy
Irish regulator probes X after Grok allegedly generated sexual images of children
Ireland’s Data Protection Commission opened a probe into X over Grok AI tool allegedly generating sexual images, including of children. Ireland’s Data Protection Commission has launched another investigation into X over Grok’s AI image generator. The probe focuses on reports that the tool created large volumes of non-consensual and sexualized images, including content involving children,…