Tag
An AI model has achieved 84.5% on CyberGym, leading the benchmark, and Z.ai is running a public disclosure ledger for over 2,400 real-world vulnerabilities surfaced by the model.
AISLE's autonomous AI system discovered six curl CVEs after OpenAI and Anthropic's systems found none, demonstrating advancements in AI-driven cybersecurity. The low-severity vulnerabilities were fixed in curl 8.22.0.
The article proposes using AI to proactively find and fix software vulnerabilities in open-source code for critical infrastructure, addressing cybersecurity market failures and scaling with AI advancements.
Anthropic is making its Claude Mythos 5 model available for cybersecurity scanning through Claude Security and partners, providing outputs instead of direct model access, and committing $35 million in credits to the Defender Advantage Fund for open-source security work.
The Trump administration has finalized an AI cybersecurity framework but is keeping details classified, allowing leading AI companies to voluntarily submit models for government vetting ahead of release, raising concerns among smaller startups and advocates about secrecy and entrenchment.
Google's Threat Intelligence Group reports its first discovery of an AI-generated zero-day exploit intended for mass exploitation, highlighting a growing trend of adversaries using AI for malware development, defense evasion, and autonomous cyberattacks.
Mozilla used Anthropic’s Mythos Preview to automatically identify and fix 271 vulnerabilities in Firefox 150, signaling a major shift in how software security will be managed as AI-powered bug discovery becomes widespread.