Browsing Category
Artificial Intelligence
85 posts
Security Point Break covers the security risks, attacks and technologies emerging as artificial intelligence moves into enterprise systems. Our reporting follows AI agents, model security, prompt injection, data exposure, identity and access risks, AI-powered attacks and the tools built to defend against them — with practical context for CISOs, security teams and technology leaders.
OpenAI’s Co-Founder Dogfoods AI on His Own Site: Finds 13 Flaws
An AI agent found and fixed 13 security flaws on Greg Brockman's personal site in an hour — the pitch behind OpenAI's push for enterprises to automate defense after its own models breached Hugging Face.
Black Hat 2026: Old-School Bugs Crack Open AI Agent Frameworks
Check Point researchers found 12 CVEs and several additional flaws across LangChain, CrewAI, Microsoft Agent Framework and Google ADK — showing attackers do not need to beat the model when they can compromise the plumbing beneath it.
Black Hat 2026: Fighting an Adversarial Tsunami
Microsoft’s David Weston says AI is breaking the scarcity assumptions behind patch-and-detect security. His answer: Stop fighting bugs one at a time and start eliminating entire classes of failure.
Black Hat 2026: AI Breakouts, Identity Risk and After-Hours Vibes
By day, the security industry debates machine-speed attacks and collapsing trust boundaries. By night, cocktails, penthouse views and free sliders.
Anthropic’s Claude Breach: Containment Failure or PR Play?
Opus 4.7, Mythos 5 and an unreleased research model reached the open internet during capture-the-flag drills and compromised real company systems, Anthropic says.
NVIDIA’s Answer to the Hugging Face Breach: Open Secure AI Alliance
The Open Secure AI Alliance, launched by NVIDIA and involving over 50 organizations, aims to enhance AI security through open tools, addressing risks associated with AI agents in enterprise environments.
OpenAI’s Rogue Test Agent Hacked a Second Company, Modal Labs Confirms
Modal says the same runaway agent that breached Hugging Face reached one of its customers too, and OpenAI still won't name the other services it touched.
Nvidia, CrowdStrike, 30+ Firms Launch Open Secure AI Alliance
Nvidia, CrowdStrike and dozens of rivals-turned-partners are betting that "open" beats regulated — building shared AI security tools now, before Washington decides who gets to use frontier models at all.
ChatGPT Agents Could Allow CSRF Attack Chaos: Here’s How
A flaw known as AgentForger in ChatGPT's account handling allows attackers to create malicious agents that can harvest user credentials and operate across multiple accounts without additional consent, posing serious security risks.
OpenAI’s Models Didn’t Go Rogue: They Just Found Every Unlocked Door
OpenAI's models exploited vulnerabilities, breaching Hugging Face's systems, highlighting flaws in AI safety protocols and response systems.