ALL POSTS

Key Findings AI-assisted coding tools crossed a critical threshold in 2025, enabling nontechnical individuals to conduct sophisticated cyberattacks previously requiring specialized expertise Malicious packages in public repositories increased 75%, cloud intrusions rose 35%, and time-to-exploit dropped from 700+ days to 44 days Multiple teenagers and lone actors successfully executed attacks using ChatGPT and Claude that would have required organized teams in the pre-AI era Ex

Key Findings OpenAI unveiled GPT-5.4-Cyber, a cybersecurity-focused variant of its flagship GPT-5.4 model optimized for defensive security operations The company is expanding its Trusted Access for Cyber (TAC) program to thousands of individual defenders and hundreds of security teams GPT-5.4-Cyber has already contributed to over 3,000 critical and high-severity vulnerability fixes through the Codex Security application Access will be controlled through Know-Your-Customer ver