A recent article compares a new AI surveillance technique to the fictional Batman technology in *The Dark Knight*. The piece was prompted by an Instagram post from O23 EcoTech showing how researchers
The UK Artificial Intelligence Security Institute (UK AISI) and the U.S. Center for AI Standards and Innovation (CAISI) jointly evaluated Moonshot AI’s Kimi K3 model for cyber capabilities shortly aft
The Trump administration has launched “Gold Eagle,” a new federal clearinghouse designed to share AI-related cyber threat information between government and private-sector partners. Created by White H
exploitbench.ai
news
exploitbench.ai
IrregularChat: Purple Team
Jun 4
ExploitBench v8-bench introduces a new way to evaluate AI on software exploitation: not as a simple pass/fail task, but as a ladder of capabilities from reaching vulnerable code to triggering the bug,
Researchers at Dartmouth College have developed a way to improve and direct Wi‑Fi signals using a 3D-printed reflector covered in aluminum foil. The technique works by exploiting a basic property of e
Researchers and tech experts say aluminum foil behind a Wi‑Fi router can sometimes improve signal, but only in a limited, directional way. A 2017 Dartmouth College study found that a properly curved r
A recent study warns that reliance on AI for cognitive tasks—like writing, studying, and coding—can impair intellectual abilities and persistence. Researchers found that while AI assistance initially
AEG: Automatic Exploit Generation
news
ndss-symposium.org
IrregularChat: Tech
Apr 11
AEG (Automatic Exploit Generation) is the first end-to-end system that automatically identifies vulnerabilities in programs and generates corresponding exploits. In testing 14 open-source projects, AE
The 2026 Artificial Intelligence Workshop, organized by MORS, will take place from April 20-23, 2026, at Carnegie Mellon University and RAND in Pittsburgh, PA. This four-day event aims to enhance know
AI Models on Realistic Cyber Ranges \ red.anthropic.com
news
red.anthropic.com
fnFo7CqDg0hZ6zASBF+Z1CdjGWMAtxfk6jKLmz2LdtY=
Jan 22
Recent evaluations reveal significant advancements in AI models' capabilities for cybersecurity, particularly the Claude Sonnet 4.5 model. Unlike its predecessor, Sonnet 3.5, which required specialize