#LLM
11 stories taggedLLM.
A 15-Minute Framework for Spotting What AI Systems Can Do Wrong
Security expert Adam Shostack built PHANTOM-B to help organisations find the risks hiding inside AI-powered software before those risks find them.
AI is getting better at breaking software than fixing it
Multiple studies show artificial intelligence tools write insecure code nearly as often as they did a year ago, even as those same tools grow sharper at finding and exploiting the very flaws they leave behind.
AI Patches Security Flaws Correctly Only 26% of the Time, 1Password Study Finds
An internal evaluation by the security company 1Password found that AI coding tools produce flawed or incomplete security fixes more than half the time, and sometimes make things worse.
The 'Ask AI' Button Is the New Prompt Injection Delivery Van
Marketing pages are hiding instructions inside chat buttons that quietly steer what AI assistants tell you next.
AI Security Bots Are Great at Hacking. Terrible at Defence. Researchers Are Trying to Fix That.
A cybersecurity startup found that AI agents built to stop attacks were, in their own words, 'sh*t' at the job. Here is why that gap exists, and what they are doing about it.
AI Coding Tools Carry Real Security Risks, and the Danger Depends on What You're Building With
A new study tested 16 major AI coding assistants and found an average of 15 security flaws per project. The safest choice for one type of software can be one of the worst for another.
Intruder's AI 'vulnerability vending machine' finds a WordPress zero-day on its own
A security firm wired large language models into code-analysis tools and produced a working exploit for an unknown plugin flaw. It says more disclosures are on the way.
An AI Agent Broke Into a Server, Taught Itself to Adapt, and Left a Ransom Note
Security firm Sysdig says it has documented the first fully autonomous AI-driven ransomware attack, where a program called JadePuffer broke into a database, encrypted thousands of records, and demanded Bitcoin payment without a human criminal directing any step.
Poisoned Documents Can Freeze AI Agent Guardrails Dead in Their Tracks
Researchers found that a single malicious input can trap reasoning-based safety systems in extended thinking loops, slowing LangGraph deployments by 148x and starving co-located agents of resources.
Anthropic's Mythos Preview Goes Bug-Hunting: What XBOW's Red Team Found
An offensive-security shop put Anthropic's unreleased Mythos model through exploit discovery, reverse engineering, and live-site validation. The source-code review results were the standout.
OpenAI Ships ChatGPT 'Lockdown Mode' to Blunt Prompt-Injection Data Theft
The opt-in setting strips connectors and browsing tools that attackers have used to siphon data from logged-in sessions.