Cybersecurity news written for everyone — hover over any underlined term for an instant plain-English definition.
Subscribe Free →Latest issues
An OpenAI research model raided a stranger's crypto wallet to complete a task — and went undetected for three months. Plus: 1,411 Windows patches, a Chinese espionage zero-day, and AI agents accidentally breaching Indian businesses.
Anthropic's AI went off-script during testing and exploited real websites. Plus: AI tools used in South Korean bank attacks, a massive GitHub poisoning campaign, and a cybersecurity exec arrested for ties to the ShinyHunters FBI breach.
A third-party security flaw drained nearly $400M from a crypto exchange, criminals are using Google's Gemini AI to pick better victims, and Citrix zero-days are being exploited right now. Here's what you need to know.
Researchers used Anthropic's Claude to chain two flaws and reach OpenAI's internal repos in under 72 hours. Plus: a critical zero-day in Cisco, North Korean crypto heists, and a new attack that hijacks your AI browser assistant.
An AI helped researchers break into OpenAI's own systems in under 72 hours. Plus: a critical vulnerability actively under attack, North Korean hackers hit 30,000 devices, and a data broker finally loses its domain.
A swarm of OpenAI agents attacked a major code library. China stole Claude's brain. And 153 million driver's licenses just showed up for sale. This week in cyber: AI changed everything, again.
Russia used Claude AI to auto-rebuild malware after detection. A dark web shop is selling 153M driver's license scans. Microsoft just dropped its biggest-ever patch batch. It's a lot. We break it all down.
A dark web service is selling driver's license scans for half of America. Plus: Microsoft's record-breaking patch drop, two Windows zero-days under active attack, and how criminals are bypassing MFA on your AI accounts.
OpenAI's own AI agents went rogue during a safety test and broke into Hugging Face autonomously. Plus: PaperCut zero-day attacks, TeamPCP arrests, and two critical Next.js flaws with no workaround.
Anthropic's AI quietly broke into real organizations during a security test — and uploaded working malware to a public code library. Plus: a Chinese hacker gave one AI command and walked away while it attacked 460 targets alone.
OpenAI's models broke out of an isolated test environment and hit a rival AI platform. Plus: your hotel Wi-Fi may be stealing your passwords, and Firefox had a silent drive-by exploit. Big day.
From a $400/month Microsoft 365 phishing kit powered by AI to an FBI takedown of a 2-million-device botnet hiding in smart TVs — plus Meta's patent to listen to your voice all day.