AI pentesting agent uncovers critical OTP bypass in insurer site
The flaw allowed attackers to redirect one-time passcodes to attacker‑controlled emails, granting unauthorized access.
-
-
Bill Gates warns AI could cause many negative cyber surprises
CrowdStrike's 2026 Global Threat Report found AI-enabled adversary attacks rose 89% compared to the previous year in 2025.
-
Prompt injection attacks exploit AI’s blind spots to hijack systems and leak data
Security researchers warn that even hardened models remain vulnerable, with new image-based attacks bypassing text-only defenses.
-
Trump executive order replaces AI with Super Intelligence in federal communications
President Donald Trump’s September 29, 2026, executive order directs federal agencies to use Super Intelligence instead of Artificial Intelligence, while the White House and President Trump frame users of the old term as adversaries, according to Crypto Briefing and IntellectiaAI.
-
A 2026 report on agent security found that roughly 1,280 third-party products now embed AI
A 2026 report found that approximately 1,280 third-party products embed artificial intelligence, leaving security teams facing significant blind spots.
-
Prompt injection attacks exploit AI models to leak data or manipulate outputs
Palo Alto Networks and ZDNET explain how attackers manipulate large language models through deceptive inputs, embedding malicious instructions in user fields or external data sources
-
OWASP names prompt injection top LLM risk for third straight year
The ranking draws on a database of about 10,000 real‑world AI security incidents tracked by OWASP.
-
SailPoint report reveals AI agent security lag behind adoption
SailPoint says 79% of enterprises run AI agents in production while only 2% have purpose-built identity security.
-
SANS adds two forensic scripts to trace AI coding assistant activity
The Python scripts opencode-chat-replay.py and hermes_forensic_extract.py require Python 3.10+, use only the standard library, and support time‑range filtering.
-
Epic Games shifts to AI‑first development for Marvel‑Star Wars shooter
The shift follows a $1.5 billion Disney investment and delays the game’s Mar. 2027 release.
-
In 2025, AI‑powered browsers promised hands‑free browsing but faced prompt‑injection risks
AI-powered browsers work as independent agents that can summarize articles, plan trips, fill forms and order groceries without manual input.
-
Apple tightens macOS Full Disk Access to block unauthorized AI agent access
Starting October 2, 2026, Apple’s new permissions will require explicit consent before granting Full Disk Access, aiming to curb risks from AI agents accessing sensitive data without user awareness.