AI Briefing#cybersecurity
The latest AI and developer news about #cybersecurity, with the original source and a short summary.
The latest AI and developer news about #cybersecurity, with the original source and a short summary.
Introducing LangSmith Sandboxes, which provides hardware-level isolation so AI agents can safely and independently execute code.
Anthropic simultaneously released Claude Fable 5, a highly capable general-purpose model, and Mythos 5, a version with relaxed security restrictions.
Apple Intelligence introduces an agentic feature that automatically changes leaked or vulnerable passwords by navigating websites.
Google Security Operations combines with AI Threat Defense to automatically detect and respond to advanced AI-driven security threats.
Fake shopping sites impersonating well-known brands were discovered in ChatGPT search results, raising concerns about the risk of users' financial information being leaked.
A bug in Instagram's support chatbot put around 20,000 user accounts at risk of having their information leaked.
Microsoft's open source projects were hacked to distribute malware that steals credentials from AI development tool users.
A new supply chain security threat has been discovered that exploits configuration files in AI coding tools such as VS Code, Cursor, and Claude Code.
Anthropic has deployed engineers to the NSA to deploy **Mythos**, a model for offensive cyber operations.
OpenAI has launched 'Lockdown Mode' to block data exfiltration through prompt injection attacks.
Successfully demonstrated a proof-of-concept for an autonomous AI worm that formulates its own attack strategies and propagates using open-weight LLMs.
Anthropic has released an open source reference framework that uses Claude to automatically detect and fix software vulnerabilities.
An open-source browser extension that defends AI agents against prompt injection and dark pattern attacks has been launched.
University of Toronto researchers demonstrated the danger of an 'AI worm' that modifies its own strategy and spreads using publicly available AI models.
Anthropic is expanding its Mythos model, specialized in detecting security vulnerabilities, to 150 companies across 15 countries.
President Trump signed a scaled-back AI executive order centered on voluntary security reviews before the release of powerful AI models.
Anthropic is expanding Project Glasswing partners to 150 organizations to strengthen software security.
New LLM attack patterns have emerged that go beyond simple instruction-ignoring, leveraging multi-message setups and frame redefinition.
A new benchmark called CVE-Bench has been released to evaluate LLM agents' ability to patch security vulnerabilities.
Google Cloud announced major AI updates including Gemini 3.5, Gemini Omni, and its AI security solution Google AI Threat Defense.
NVIDIA unveiled its RTX Spark superchip, and the first case of an autonomous LLM agent cyberattack has been reported.
A study has found that when LLM agents fix security bugs, the most dangerous outcome is a 'plausible but wrong patch' that passes tests while leaving the vulnerability in place.
In EY Canada's cybersecurity report, the majority of citations turned out to be AI hallucinations and false information.
PolyRange, a new cybersecurity AI evaluation methodology that solves the data contamination problem through an LLM-generation approach, has been released.