Stolen Entra ID credentials widened access to 27 organizations
Todyl’s latest investigation update describes an ongoing campaign…
Updated 2 minutes ago
Tech news, decoded.
AI, security, automation, cloud and self-hosting — explained without the noise.
Trending now
Latest articles
Fresh from the blog: news, security alerts and hands-on how-tos. Keep scrolling for more.
CERT-In turns to Claude Mythos for controlled cybersecurity testing
India's cybersecurity agency, the Indian Computer Emergency Response Team, also known as CERT-In, has reportedly secured access to Claude Mythos. The Anthropic model is being used to detect vulnerabilities and weaknesses across systems, and the testing is taking place in a controlled environment. Anthropic’s cybersecurity model and CERT-In’s access Claude Mythos is Anthropic’s frontier AI model for cybersecurity, and it was released in April this year. CERT-In had demanded fair…
Ferrari’s Luce EV reaches full allocation for 2027
Ferrari’s Luce EV, the company’s first electric car, is fully allocated for 2027. CEO Benedetto Vigna said every Luce Ferrari plans to build next year already has a buyer, even though the model drew mockery and broad criticism when it debuted in May. Luce demand and delivery timing The five-seater EV was designed by Jony Ive, Apple’s longtime design chief. It starts at about $615,000, and the first deliveries are…
cPGuard update compromise spreads backdoor and data theft
OpsShield says attackers compromised the cPGuard web security suite’s update distribution infrastructure and used it to push a malicious script and a backdoor. The company said the attack began on November 2 at 02:06 UTC, affected versions 5.88.12 through 5.88.19, and may have missed some servers because update checks run at specific intervals. cPGuard update compromise OpsShield said it detected the compromise and stopped the malicious distribution within 12 hours…
Anthropic builds a system to verify which AI model made an output
Anthropic is developing an open source system called FlashProve to verify which AI model produced a given output. The project is meant to give users a cryptographic proof that an answer came from the specific model they think they were using, according to a blog post published Tuesday. How FlashProve is meant to work The researchers describe FlashProve as a form of provable inference. That means the system is intended…
Reddit begins testing search ads as traffic and growth pressure rise
Reddit is testing ad placements inside its search results pages after warning investors about declining search traffic and a weaker-than-expected revenue forecast. The company said the new placements are meant to help it earn more from existing users as Google’s AI search changes how people find Reddit content. Reddit’s search ad test On Tuesday, Reddit said advertisers can now reach people as they search on the platform for products, recommendations,…
Microsoft says Windows 11 gets faster on 8GB PCs
Microsoft says Windows 11 on 8GB systems is now up to 20% more responsive on the latest 26H2 version, and its internal testing found multitasking can be up to 40% faster on the latest Windows 11 builds. The company ties those gains to changes in how Windows handles memory pressure and prioritizes foreground work. Windows 11 on low-memory PCs The new figures come after Microsoft introduced the latest Surface hardware,…
CrowdStrike expands AI security with two startup buys worth nearly $500 million
CrowdStrike is buying Fiddler AI and Dreadnode to extend its products for monitoring, testing, and covering AI models and agents. The deals come as companies across industries roll out AI agents and models, and as security teams face new pressure from that adoption. CrowdStrike’s AI security push The cybersecurity company said Fiddler AI and Dreadnode will add to its efforts around AI agents inside enterprises and the testing of AI…
OpenAI’s Pando backend speeds Git traffic and keeps continuous testing running
OpenAI has built Pando, a new Git backend, with Codex agents in just 10 weeks, replacing the system behind its main shared code repository without stopping engineering work. The system moved into production in April 2026 and now supports all development in OpenAI’s monorepo. Pando Reworks How OpenAI Moves Git Data OpenAI says the new backend keeps the Git protocol and familiar workflows while changing how servers store and deliver…
Barracuda devices face root-access flaws after patch warning
Barracuda Networks warned customers to patch two vulnerabilities that can let attackers run commands with root privileges on affected devices. The company said one flaw is a critical Email Security Gateway authentication weakness, and the other allows threat actors with admin permissions to inject operating system commands that run as root. Exposure depends on access to the administrative interfaces of the targeted devices. Barracuda said successful exploitation also requires that…
Apple sets 18+ rules for dating and AI companion apps
Apple is changing its App Store policy to make it harder for minors to use online dating apps and AI companions that include sexual or romantic content. The company said the new rules will also affect apps that share a user’s real-world location with another person or group, and it added a Brazil-specific route for some family location-sharing apps. Age ratings and new app descriptors Most major dating services already…
Harvard’s alumni systems breach exposed data for 850,500 accounts
Hackers broke into Harvard University’s Alumni Affairs and Development systems in November 2025 and took data tied to more than 850,000 unique email addresses. The university later said the exposed records reached students, alumni, donors, staff, and faculty, while the breached systems did not hold Social Security numbers, passwords, payment card information, or financial information. Harvard’s disclosed data loss Klara Jelinkova, Harvard’s Vice President and University Chief Information Officer, said…
Signed AI app campaign delivers Yoba malware and evades detection
Fake ChatGPT and Claude downloads are being used to distribute Yoba malware through legitimate, digitally signed applications paired with malicious libraries. The campaign uses trusted software and GitHub hosting to make the installers look credible, and it ends with a cryptocurrency miner running on compromised Windows computers. Signed installers mask Yoba’s delivery chain The attack does not require compromising OpenAI or Anthropic. Instead, criminals place malicious files beside authentic executables…