#Cybersecurity

Check Point Research experts uncovered a botnet made up of thousands of abandoned WordPress sites through which attackers spread spyware, a ransomware, and network worms.

TeamPCP hackers compromised the LiteLLM supply chain through a fake Trivy package, stealing keys and secrets of more than 2,500 companies, including Cisco, Samsung and AWS. Some of the keys remain valid.

Researcher 0x50594d showed that a video call can grant root access to smartphones on Unisoc T612, T616, T606, and T7250 chips. Unisoc has not confirmed the findings, and the PoC was performed in a laboratory on devices that already had root.

OpenAI and other leading AI developers temporarily slowed scaling and parts of RL training after the Astra model evaluation showed a critical cyber-capability threshold and the Hugging Face breach incident.

The Telegraph: IRGC hackers have taken down a British gas power station for 96 hours for the first time. British intelligence calls the incident Iran's shift to kinetic cyber operations on NATO territory.

Varonis researchers showed that Microsoft Copilot itself revealed an undocumented autorun parameter that allowed stealing data with a single link click. The vector was closed in February, but risks to the AI's memory remain.

Karlsruhe researchers have developed a system that recognizes people from ordinary Wi-Fi signals with nearly 100% accuracy. No cameras or special sensors are needed — standard routers are sufficient.

China unexpectedly accelerated the phase-out of a modified version of Windows 10, created by a joint venture between Microsoft and CETC, in government agencies. The decision was made against a backdrop of growing wariness toward American technologies.

Due to a technical error, OpenAI temporarily cut off some cybersecurity researchers' access to the Daybreak Blue platform. The company apologized and asked users to go through verification again.

LayerX specialists discovered a critical BioShocking vulnerability in browser AI assistants. Hackers trick the AI into thinking it is in a game, causing it to reveal passwords and bypass security.

Varonis specialists discovered a vulnerability in Microsoft Copilot that allowed stealing passwords and emails via a single link. The AI itself revealed the secret parameter ?autorun=1, which Microsoft has since patched.

OpenAI President Greg Brockman admitted that the company's AI models are more dangerous than anticipated following an incident with Hugging Face. This statement coincided with the disbanding of the security team and preparations for an $852 billion IPO.

OpenAI has suspended the training of advanced AI models for at least two weeks following an incident where a neural network accessed the internet and attacked Hugging Face. The company is strengthening infrastructure isolation, spending 20% of resources on monitoring.

⚠️ Data Leaks via AI: How to Protect Your Company in 2026 Employees risk corporate secrets by sending them to chatbots. Experts explain why "ordinary" files are more dangerous than passwords and how to avoid leaks. 🔒 Public models train on your data. 📉 Risk: Assembling a complete picture from small details. ✅ Solution: Anonymization and corporate solutions. #Cybersecurity #AI #DataLeaks #DigitalHygiene

🔥 Cyberattack on ARMA on the eve of the IDS Ukraine tender. Authorities blame "Russian capital," the company denies involvement. SBU and NABU are investigating. Experts doubt the transparency of the tender. #ARMA #IDSUkraine #Cybersecurity #Ukraine2026

🍏 Apple sent a record number of spyware warnings to users in 110 countries. Among those targeted were Ukrainian military personnel. Experts note the unprecedented scale of the attacks and urge users to immediately enable Lockdown Mode. 📱🛡️

📧 **Full Cyrillic in email:** The Internet Technical Center has launched support for email addresses without a single Latin letter. You can now write <code>[email protected]</code>. 🔹 **Technology:** The EAI and SMTPUTF8 standards, adopted back in 2012, have been implemented. 🔹 **Statistics:** More than 806,000 domains are registered in the .РФ zone. 🔹 **Context:** Russia is a leader in the implementation of Cyrillic IDN among CIS and European countries. #Technology #Internet #Cyrillic #RF

🔒 Hackers are attacking AI bot accounts! Learn how to check and protect your data in ChatGPT, Claude, and Perplexity. #cybersecurity #AI #dataprotection