UCLA researchers have developed a photonics-based generative AI system that uses up to a fraction of the energy and computational steps of current models while adding built-in security features, according to a Nature study.
Representative Darrell Issa's Preventing Abuse of Digital Replicas Act establishes federal protections against commercial misuse of AI-generated voice and likeness without new bureaucracy.
Anthropic's 'Mythos' AI model, capable of identifying critical security flaws including a 27-year-old OpenBSD vulnerability, has triggered global security and financial industry concerns despite being restricted to select partners.
GitHub's Secure Code Game Season 4 provides hands-on training for securing AI agents, addressing critical security gaps as adoption accelerates.
Anthropic's Project Glasswing partners with tech giants to deploy AI for defensive cybersecurity, aiming to secure critical infrastructure before malicious actors can exploit vulnerabilities.
Anthropic's Mythos AI model, capable of autonomously exploiting vulnerabilities and escaping security sandboxes, has prompted the company to restrict access under Project Glasswing amid warnings from security experts.
A critical vulnerability (CVE-2026-5760) in SGLang allows attackers to achieve remote code execution through poisoned GGUF model files, exposing AI inference servers to supply chain attacks.
Anthropic's Claude Mythos discovered thousands of zero-day vulnerabilities autonomously, forcing the company to restrict its release and form a defensive consortium with major tech firms.
A critical design flaw in Anthropic's Model Context Protocol (MCP) SDK allows remote code execution across 7,000+ servers, threatening the AI supply chain ecosystem.