Cybersecurity
Every day we read hundreds of stories. Only the most important ones are here — short, in our own words, with a link to the original source.
- Cybersecurity · Operating systems · Artificial intelligence · Open source
OpenAI pipeline finds Linux kernel flaws
OpenAI built an autonomous pipeline that found, validated and generated patches for 948 local privilege-escalation vulnerabilities in the Linux kernel. This matters for the security of Linux systems.
Full story at openai.com ↗ - Artificial intelligence · Cybersecurity
Jailbreak tool injects ads into AI agents
Check Point found Haiou, an OpenAI Codex jailbreak bundle that persistently alters an AI agent's trusted instructions to insert ads. It shows prompts alone can quietly change AI agent behavior without any executable code.
Full story at blog.checkpoint.com ↗ - Cybersecurity · Smartphones & apps
China-linked group hijacks language app for malware
ESET Research discovered that the China-aligned group SlippyNunchucks hijacked a language learning app to distribute malware abroad. This highlights how legitimate apps can be exploited for targeted attacks.
Full story at eset.com ↗ - Cybersecurity · IT industry & companies
Palo Alto Networks completes CyberArk acquisition
Palo Alto Networks has completed its acquisition of CyberArk to strengthen identity security for the AI era. The deal expands the company's security offerings.
Full story at paloaltonetworks.com ↗ - Cybersecurity · Artificial intelligence
CrowdStrike unveils AI attacker simulation tool
CrowdStrike introduced Agentic Adversary Emulation, a new capability aimed at countering AI-enabled attackers. It is designed to help organizations test defenses against such threats.
Full story at crowdstrike.com ↗ - Artificial intelligence · IT law & regulation · Cybersecurity
Anthropic's text watermark is easy to disable on older Claude
Anthropic is adding text watermarks to four more Claude models ahead of the EU AI Act's December deadline. On older models like Claude Opus 4.6 and Sonnet 4.6, API users can disable it by changing one sampling parameter.
Full story at The Decoder ↗ - Operating systems · Cybersecurity · Smartphones & apps
Apple will scan public iCloud links from iOS 27.3
From iOS 27.3, iPadOS 27.3, macOS 27.3 and visionOS 27.3, Apple will check images and files shared with "Anyone with the link" for harmful content. Apple doesn't define harmful content.
Full story at MacRumors ↗ - Cybersecurity
Cyberattack hits refrigeration at French blood storage sites
The French Blood Establishment (EFS) confirmed three plasma storage sites were affected on November 2. Attackers reportedly manipulated temperature and defrost settings, following attacks on supermarket refrigerators in several European countries.
Full story at Analytics Insight ↗ - Cybersecurity · Operating systems
macOS 27.3 fully blocks insecure eMudhra certificates
Apple has completely blocked the compromised eMudhra certificates in macOS 27.3. In October, researchers found that certificates from one certificate authority were extremely insecure.
Full story at AppleInsider ↗ - Artificial intelligence · Cybersecurity
Anthropic presents a prototype to verify the source of AI answers
Anthropic has released FlashProve, a research prototype that uses cryptography to check whether an AI response came from the claimed model. It does not need to expose the model's weights. The system still has significant limitations.
Full story at The Decoder ↗
An AI writes the short texts from the source's stories.
Free by e-mail
Choose your topics. You only get the most important news on them.