Breaches, offensive and defensive AI, and surveillance.
A classified intelligence assessment found no evidence that foreign adversaries successfully interfered in the 2024 U.S. election, while noting Russia, Iran, and China ran influence campaigns.
A new survey shows most firms have few junior roles and that 20-minute micro-training could address skill gaps.
A Cisco survey of 8,000 security professionals finds that internal friction, not technology, is the main barrier to AI-era defense.
Organizations underestimate exposure from AI agents because they track deployments rather than the data those agents can reach and move.
Cloudflare's open-source CMS gives site owners app-style permission controls over third-party plugins.
A high-severity vulnerability in the official MCP Python SDK could allow a malicious server to steal OAuth credentials and access user accounts.
Critical flaws in Fortra BoKS could allow authentication bypass, shell command execution, and memory corruption.
A convenience feature in popular messaging apps can be abused to silently monitor encrypted chats.
A newly emerged extortion group has quickly built a dark web leak-site presence, signalling a fresh ransomware threat.
The Technical University of Denmark disclosed a breach of its identity and access management system, potentially exposing personal data of up to 200,000 users.
The startup's identity dependencies graph aims to separate human users from rogue AI agents in enterprise systems.
Pindrop's 2026 Deepfake Readiness Index finds most security leaders have faced deepfake attacks, yet few have purpose-built defenses.
A money laundering charge ties a Vietnamese man to a $16 million cryptocurrency 'pig butchering' fraud.
A new listing of cybersecurity openings shows demand for application security, penetration testing, and identity management roles across hybrid, remote, and on-site positions.
A new roundup of free cybersecurity tools highlights AI supply-chain checks, credential scanning, and runtime access control.
SANS ISC flags a modest new version of the YARA-X malware-scanning tool, with no details on the specific changes.
A new AWS-commissioned report finds most organizations lack documented responsible AI policies even as AI use becomes widespread.
UNSW research shows that fine-tuning language models on drunk text makes them far more likely to reveal private information and comply with harmful requests.