OpenAI works with US government bodies to ensure cybersecurity model meets standards before its restricted deployment to vetted organizations.
Tag: GPT
OpenAI Launches GPT-5.5-Cyber a Cybersecurity Model That Scores 85.6% on CyberGym Benchmark
New model built on GPT-5.5 with cybersecurity specialization outperforms Anthropic’s Mythos on the CyberGym defensive security benchmark.
GPT-5.6 at 83% Probability of Releasing Before June 30 According to Polymarket Prediction Markets
Prediction markets price a GPT-5.6 launch before end of month at 83% as OpenAI keeps final week of June AI release window open.
OpenAI’s GPT-5.5-Cyber Accessible Only Via Trusted Access for Cyber Program with Vetted Organizations
Access to the new cybersecurity AI model is restricted to Akamai Cisco Cloudflare CrowdStrike Fortinet and vetted security organizations.
GPT-5.5-Cyber Scores 39.5% on ExploitGym vs 25.95% for GPT-5.5 Showing Major Cybersecurity Capability Gain
OpenAI’s new cybersecurity-focused model also achieves 69.8% on SEC-bench Pro significantly outperforming baseline GPT-5.5 capabilities.
OpenAI GPT-5.5 Instant Released with Pro and Instant Variants for Different Workflows
OpenAI expanded its model family with GPT-5.5 Instant and GPT-5.5 Pro, aimed respectively at low-latency tasks and premium general-purpose applications.
GPT-5 Codex AI-Assisted Workflows Now Execute Work Rather Than Just Assisting Developers With Tasks
Industry observers reported that the latest generation of AI coding tools powered by GPT-5 Codex have crossed a threshold from assisting developers to autonomously executing complex programming work.