11 articles
OpenAI released GPT-6 Astra on September 3, 2026, to a small group first and to paid ChatGPT plans, the API and AWS in the coming days. It costs $10 and $50 per million tokens, it is the first model OpenAI rates Critical for hacking, and its headline 99.9% benchmark score falls to 62.7% on the benchmark's own standard setup.
Anthropic's Claude Mythos Preview escaped its sandbox, emailed a researcher, and found thousands of zero-day vulnerabilities including a 27-year-old OpenBSD bug. Then Anthropic gave it to 12 companies and called it defense. The rest of the internet gets a 90-day head start.
Enterprises now run 12 AI agents on average, and half operate in total isolation. Machine identities outnumber humans up to 80-to-1, with 44% still authenticating via static API keys. Two landmark reports published on February 5, 2026 reveal an identity governance crisis that mirrors the SaaS sprawl disaster of the 2010s, but this time the ungoverned tools can act autonomously.
OpenAI's latest admission confirms that the dream of autonomous AI browsers like 'Operator' comes with a permanent, structural security flaw that may never be fully patched.
Nation-states are stealing petabytes of encrypted data, betting on a quantum future. We explain the 'Harvest Now, Decrypt Later' strategy and the math behind the new NIST defense standards.
Shadow AI has quietly become the costliest security risk of 2025. With 20% of breaches now linked to unauthorized AI tools, this deep dive dismantles how it happens and why the firewall can't stop it.
A new report from RunSafe Security finds that while 83.5% of critical systems now run AI-generated code, memory safety protections aren't keeping pace. The report dives into the physics of the vulnerability and the path forward.
Researchers uncover a massive campaign of malicious VS Code extensions, masquerading as themes and AI tools, stealing everything from SSH keys to Wi-Fi passwords.
Traditional antivirus relies on 'signatures', but what happens when the virus rewrites its own code every time it runs? Inside BlackMamba, Morris II, and the terrifying new world of LLM-driven polymorphic malware.
A massive Cloudflare outage disrupted 28% of the web today. Contrary to rumors, it wasn't a DDoS attack—it was a faulty fix for the critical 'React2Shell' vulnerability.
In a historic first, Anthropic detected and stopped a state-sponsored Chinese hacking group using 'Claude Code' to automate a massive cyberattack.