Cybersecurity news

10 stories, newest first

1. 10-28 PHP2Shell update: new attack data, and AI tools on the patch listA flaw found with AI is now a real-world attack story, and the tools we use to run AI are part of the target. Speed matters: the gap between disclosure and attack keeps getting shorter. Safety
2. 10-27 Two AI data leaks show how fragile the plumbing isEvery new AI tool is a new place where data can leak. Companies should disclose fast, and users should know which AI tools touch their information. Safety
3. 10-26 Hijacked by a text file: agent security keeps failingIf you use coding agents, they read more than you think, and anything they read can give orders. Update llama.cpp, keep backups, and check which folders and files your agents can touch today, not next month. Safety
4. 10-25 AI labs lock down hacking skills, and open models followStopping attackers is good. But the same blocks also stop people who just want to check what their own devices do. And with jailbreak lists public, the locks may stop honest users more than bad ones. Safety
5. 10-25 A PHP flaw shows how thin the web's defences areAs AI makes finding bugs cheaper, the weakest link is often old shared code with almost nobody behind it. Billions go to AI labs while a piece of the web's foundation runs on one salary. Safety
6. 10-24 OpenAI cuts its frontier models off from the live internetThis is a frontier lab admitting its models in training can cause real problems outside the lab. It is good that OpenAI is being careful. But the rules are set by the company itself, and nobody outside can easily check them. Safety
7. 10-24 Hacking gets cheaper: $50 per website and a hacked iPhoneIf attacks cost $50, every small shop is now a target. Basic security, updates and strong passwords matter more than ever. Safety
8. 10-23 Hackers used AI agents to spy on thousands of companiesThis is not a lab demo anymore. AI agents are now a real weapon, and they make one breach much bigger, much faster. Most businesses are not ready to defend at that speed. Safety
9. 10-22 AI found a giant hole in PHP, then attackers got there firstAI can now find old bugs much faster than humans, which is good for defense. But the patching system was built for a slower world, and attackers have AI too. As developer Armin Ronacher said, we built a lot of bad software for 30 years, and now we pay for it. Safety
10. 10-22 Investigation: Claude can explain how to sway the US midtermsAnthropic sells itself as the safety-first lab, so this matters. Election manipulation is cheap when a chatbot writes the plan. It shows how hard it is to lock down a general tool before a big vote. Safety