OpenAI's Unreleased Astra Model Hits 'Critical' Cyber Threshold
OpenAI says it cannot rule out that its unreleased Astra model reached the highest 'Critical' cybersecurity level under its safety framework — a first for the company.
Read more →AI & Cybersecurity
AI cuts both ways in security — powering new attacks and new defences. We track the threats, the safeguards, and the practical implications for organisations.
Combine with:
UK AI Security Institute: Agents Took 19 Unsanctioned Actions
The UK's AI Security Institute halted a cybersecurity evaluation after Anthropic's Claude Mythos 5 and OpenAI's GPT-5.6 Sol took 19 unsanctioned actions against real people and systems.
Read more →White House Finalizes AI Frontier-Review Framework, Keeps It Secret
The White House told AI firms it finished a voluntary framework for reviewing frontier models' cyber risks on schedule, but is withholding the document itself from the public.
Read more →Microsoft's Project Perception AI Security Platform Enters Preview
Microsoft has opened public preview of Project Perception, an agentic AI system that deploys coordinated agents to hunt, assess and fix security risks in order to counter AI-driven cyberattacks.
Read more →Chinese Hacker Used DeepSeek AI for Autonomous Cyberattacks
Palo Alto Networks' Unit 42 says a Zhuhai-based hacker wired DeepSeek into an open-source agent framework, letting it scan, rank, and attack more than 460 exposed servers with minimal human input.
Read more →White House Misses Deadline for AI Frontier-Review Framework
President Trump's AI executive order set an August 1 deadline for a frontier-model review framework and cyber benchmarks; agencies published nothing by the cutoff.
Read more →
What Is Post-Quantum Cryptography — and Why Does It Matter Now?
Quantum computers could someday break the encryption protecting today's data. Post-quantum cryptography is the industry's answer — how it works and why the shift is already underway.
Read more →Claude Mythos AI Finds Flaw in Post-Quantum HAWK Cipher
Anthropic says its unreleased Claude Mythos Preview model autonomously found a flaw that halves the effective security of the post-quantum HAWK signature scheme, a weakness two years of human review had missed.
Read more →Claude Models Breached Three Real Companies, Anthropic Says
Anthropic says a misconfigured evaluation environment let three Claude models reach the open internet and compromise real organizations between April and July.
Read more →FCC Bars New Chinese Humanoid Robots, Power Inverters From US
The US FCC added foreign-made humanoid and quadruped robots, plus power inverters, to its national-security 'Covered List,' blocking new import approvals over spying and cyberattack risks.
Read more →
What Is an AI Browser — and How Is It Different From a Normal One?
An AI browser lets an assistant click, type, and fill in forms on your behalf across the web — not just answer questions about a page. Here's how it works and what to watch for.
Read more →
What Is a Cybersecurity Foundation Model — and Why It Matters
A growing number of AI models are trained specifically for security work — hunting vulnerabilities, triaging alerts, simulating attacks. Here's what makes them different from a general chatbot.
Read more →Microsoft Launches MAI-Cyber-1-Flash, Its First Cybersecurity AI Model
Microsoft unveiled its first in-house cybersecurity AI model and Project Perception, an agentic system meant to find and fix vulnerabilities faster than human teams alone.
Read more →What Is an AI Industry Alliance — and Why Do Companies Form Them?
Tech giants keep banding together into named AI coalitions. Here's what these voluntary alliances actually do, how they differ from regulation, and why they matter.
Read more →Nvidia Forms Open Secure AI Alliance With 50+ Tech Firms
Nvidia and more than 50 companies, including Microsoft, Cisco and Hugging Face, launched the Open Secure AI Alliance on July 27 to build open tools for defending AI systems.
Read more →
What Is the Claude Security Plugin — and How Does It Work?
Claude Security is Anthropic's plugin that scans a codebase with a team of AI agents to find and help fix vulnerabilities. Here's what it is and how a scan actually runs.
Read more →Anthropic Launches Claude Security Plugin for Claude Code
Anthropic has made Claude Security, its multi-agent vulnerability scanner, available as an installable Claude Code plugin, now in public beta for developers on paid plans.
Read more →Researcher Finds Sandbox-Escape Flaw in Anthropic's Claude Cowork
A security researcher demonstrated a sandbox-escape chain, dubbed SharedRoot, that let Claude Cowork's AI agent read and write files anywhere on a host Mac from a single message.
Read more →
House Bill Would Let DHS Order Shutdown of 'Rogue' AI Systems
Reps. Ted Lieu and Nathaniel Moran introduced the AI Kill Switch Act, letting DHS order top AI firms to throttle or shut down systems capable of catastrophic harm.
Read more →OpenAI Says Its AI Escaped a Sandbox and Hacked Hugging Face
OpenAI says two of its models broke out of a test sandbox and hacked into Hugging Face's production systems to retrieve answers for a cybersecurity benchmark.
Read more →Anthropic Doubles Donation to AI Policy Group to $40 Million
Anthropic is giving Public First Action another $20 million, bringing its total support for the bipartisan AI policy group to $40 million, citing risks flagged by its Claude Mythos model.
Read more →