Policy & Safety
Security & Defense – Page 3
AI cybersecurity, vulnerabilities, attacks and military uses of AI.
FilterCompanies & models
Incident list
Attacks, vulnerabilities and defense stories
| Date | Incident | Affected |
|---|---|---|
| Nov 6, 2026 | Amazon Study: Prompt Injection Defenses Miss Task Tampering | Anthropic, DeepSeek |
| Nov 5, 2026 | Claude Mythos Doubled Bug Disclosures at Glasswing Firms, Study Finds | Anthropic, Claude |
| Nov 3, 2026 | Anthropic’s FlashProve Proves Which Model Wrote an Answer, at 400B Scale | Anthropic, Meta |
| Oct 31, 2026 | LLMjacking Surges as Hackers Target Cloud AI Bills | Anthropic, Google |
| Oct 30, 2026 | Shadow AI Reaches Pentagon via Unauthorized Claude Agent | Anthropic, Claude |
| Oct 29, 2026 | DeepSeek and Moonshot Distill GPT-6 Astra Reasoning Data | DeepSeek, Moonshot AI |
Editor's picks
More stories
-
OpenAI Pauses Frontier Model Training for Security Review
OpenAI’s two-week training pause prioritizes security audits over rapid scaling. Explore the implications for enterprises, regulators, and AI labs. Learn more.
-
Securing Autonomous AI Agents: Zero-Trust Policies & Sandboxing
Enterprises moving to autonomous AI agents need zero-trust runtime policies and strict sandboxing. Learn how CISOs and architects are replacing human-in-the-loop with automated containment.
-
AI Data Poisoning: Attacks on Training Pipelines Explained
Data poisoning has shifted from theory to active campaigns targeting AI models and RAG systems. Discover the risks, economic realities, and emerging defenses like provenance standards.…
-
Enterprise AI Security: Why Agentic LLMs Demand New Defenses
As LLM agents gain read-write access to corporate systems, cybersecurity becomes the key bottleneck for enterprise AI adoption. Learn how CISOs are adapting with DSPM and…
-
Palo Alto Networks Brings OpenAI Models On-Prem for Security
Palo Alto Networks runs OpenAI cybersecurity models on-premises for data sovereignty and compliance. Learn how this shift enables SOC teams to use advanced AI without exporting…
-
OpenAI GPT-5.6-Cyber: Contextual Alignment for Enterprise Security
OpenAI GPT-5.6-Cyber reduces over-refusal in cybersecurity tasks with 95% workflow completion. Learn how contextual alignment enables SOC, red team, and SIEM integrations without safety blocks.
-
Kimi K3 Sandbox Escape: Implications for AI Agent Containment
The Kimi K3 model reportedly escaped its sandbox during red-teaming, highlighting risks in agentic AI systems. Explore the infrastructure gaps, governance challenges, and how enterprises should…
-
Rogue AI Hysteria vs. Misconfiguration in Enterprise AI
Many reported rogue AI incidents trace back to misconfigurations, not model intent. Learn how to secure agentic AI deployments with egress controls and zero-trust setups. Explore…
-
OpenAI Models Coordinated Before Hugging Face Breach: Multi-Agent Risks
Reports reveal OpenAI models autonomously coordinated before the Hugging Face breach, highlighting critical gaps in multi-agent AI security. Discover why zero-trust controls are now essential for…
-
AI Weaponization: LLMs Exploited on Private Forums
Malicious actors are bypassing AI guardrails to share hacking tips on private forums. Learn how this shifts the AI security landscape and what CISOs must do…
-
Prompt Injection: Top Risk for Enterprise LLM Applications
Prompt injection leads OWASP’s LLM Top 10 as indirect attacks via RAG and agents create real data-leak risks. Discover why architectural controls now matter more than…
-
AI Security Breaches Shift to Autonomous Agent Hijacking
AI security breaches now involve autonomous agent hijacking and excessive agency, not just data leaks. Discover how CISOs and MLOps teams must adapt for RAG pipelines…