Zhipu AI from China has launched GLM-5.2, a new model that reportedly rivals Anthropic’s Mythos in cybersecurity and bug detection capabilities.
The FBI’s Kinetic Cyber Range features simulated environments, including a hotel, a gas station, and a data center, designed for individuals to engage in protest scenarios.
Amazon’s CEO Andy Jassy engaged in discussions with officials about security issues related to Anthropic’s Fable 5 just prior to the implementation of new export control measures.
YouTuber and musician Benn Jordan has transitioned from synthesizer reviews to uncovering the shortcomings of our surveillance society.
President Donald Trump has enacted an executive order to establish a system of voluntary oversight for frontier AI models prior to their launch.
Meta has resolved a vulnerability that enabled hackers to hijack Instagram accounts through its AI-driven support chatbot.
A specific four-letter term triggered a security breach.
Microsoft is considering legal action against security researcher Nightmare Eclipse for disclosing vulnerabilities to the public.
While AI lacks emotions, top hackers excel at mimicking them.
Sammy Azdoufal has discovered over 1.1 million Meari cameras that can be accessed remotely. Meari provides camera technology for various brands, including Arenti, Anran, Boifun, ieGeek, and Wyze.
A groundbreaking discovery by Google’s Threat Intelligence Group reveals the first instance of a zero-day exploit created with the help of artificial intelligence.
Move over robotic vacuums—Yarbo’s cutting-edge bladed robots present a whole new level of security concern.
A newly discovered vulnerability, CVE-2026-31431, known as the “Copy Fail” exploit, poses a significant risk to most modern Linux distributions. This flaw allows attackers, regardless of their access level, to elevate their privileges to admin status while remaining undetected.
Granola, the AI-driven note-taking application, automatically sets your notes to be accessible to anyone with a link. Additionally, it enables AI training for users who are not part of an enterprise plan.
Todd McKinnon discusses the importance of identity for AI agents, the need for security in the OpenClaw landscape, and the necessity of a cautious approach in anticipation of the SaaSpocalypse.
Microsoft has addressed a security flaw in Notepad that affects Markdown files. This vulnerability could enable an attacker to execute harmful code remotely by deceiving users into clicking a link within a Markdown document.