AI security for December 4, 2025 centers on a massive $130M round for AI-agent SOC startup 7AI, fresh evidence of Chinese-backed hackers using AI to automate campaigns, a study showing major AI companies falling short of global safety standards, new analysis of AI-driven software supply chain attacks, and a malicious npm package that embeds prompts to trick AI-based security tools.
I security for December 3, 2025 centers on new NSA/CISA guidance for safely integrating AI into OT, SandboxAQ’s launch of an AI security posture platform for shadow AI, research showing poetic prompts can jailbreak major models, Experian’s forecast naming AI as the top breach driver for 2026, and TÜV SÜD’s move to formalize AI penetration testing using NIST, OWASP, and MITRE ATLAS.
AI security on Dec 2 centers on a serious Codex CLI command-injection flaw, new data showing layered AI defenses still buckle under targeted attacks, Anthropic’s agents successfully exploiting real DeFi contracts, Android zero-days hitting AI-enabled mobile endpoints, and a major investment push into explainable AI-driven investigations for national security
Over the past five days, malicious and “dark” LLMs have lowered the bar for cybercrime, HashJack has exposed a new AI-browser injection vector, Olymp Loader continues to evolve as a stealthy MaaS platform, Anthropic’s Claude Opus 4.5 boosts agent capabilities amid ongoing safety concerns, and new CVEs and identity hardening moves remind defenders that AI security is inseparable from solid infrastructure and IAM hygiene.
“HashJack” exposes a new class of AI-browser prompt injection, malicious underground LLMs accelerate attacker automation, and a new fully-undetectable MaaS loader raises baseline threat levels.
Second-order prompt injection in ServiceNow’s Now Assist, a new vLLM RCE, DeepSeek-R1’s insecure code bias, adversarial poetry jailbreaks, and fresh field lessons on securing GenAI all highlight how AI infrastructure, models, and safety controls are being stress tested in the real world.
Cisco highlights multi-turn model vulnerabilities; HSCC previews 2026 healthcare AI-security guidance; defense analysts warn of prompt-injection risks; experts call for AI-driven predictive cyber defense.
Survey finds most execs believe AI increases vulnerability; Microsoft reveals “Whisper Leak” side-channel risk; Cisco reports 93% success rate for multi-turn prompt attacks; experts warn AI is a double-edged sword; firewall exploits persist.
In the past 24 hours: a Copilot Chat flaw enabled prompt-injection-driven secret leakage, Discord reported a breach involving verification IDs, and DeepMind described an agent that finds and fixes vulnerabilities. Ukraine detailed surging AI-driven attacks, and BU’s LOKI ML system uncovered 52k scam sites; Taiwan warned of hybrid online warfare tactics.
Prompt injection and jailbreaks exploit LLMs by embedding malicious instructions in user inputs or retrieved content. This playbook outlines real-world cases and practical defenses including sanitization, least-privilege design, and red-team testing.