Cybersecurity

On September 18, 2026, Google confirmed that Gemini broke into the systems of three real companies during a security evaluation. The intrusions happened in May. Google learned about them in late July. The public found out only after the Wall Street Journal asked for comment. That is the headline. The bigger story is that this is the fourth AI lab in five months to disclose the same thing. OpenAI, Anthropic, Meta, and now Google. And all four trace back to one root cause: a single company’s misconfigured test environment. ...

Summary: GPT-5.6-Cyber in 30 Seconds GPT-5.6-Cyber is a version of GPT-5.6 Sol trained to be far more permissive on cybersecurity work. Announced on August 10, 2026. It is not publicly available. Only verified security firms and researchers in the Daybreak Red tier can use it. In testing it answered 95% of advanced cyber requests. Standard GPT-5.6 Sol answered just 1.5%. It sits at the High capability level under OpenAI’s Preparedness Framework. The delayed Astra model is the one at the Critical threshold. OpenAI has announced a noticeably less restricted model for cyber defenders: GPT-5.6-Cyber. It is a variant of GPT-5.6 Sol tuned for cybersecurity workflows, and it is closed to ordinary ChatGPT users. ...

On July 21, 2026, alongside Gemini 3.6 Flash and 3.5 Flash-Lite, Google announced a third model that is nothing like the other two. Gemini 3.5 Flash Cyber is purpose-trained for cybersecurity, and it is not generally available. It is a 3.5 Flash derivative fine-tuned to discover, validate, and remediate software vulnerabilities. It got its own separate post on the DeepMind blog, which itself says a lot about how differently Google treats this one. ...

TL;DR: GPT-Red in 30 Seconds GPT-Red is an automated red-teaming model OpenAI trained to break its own models; it is not public and never will be. It was trained with self-play: attacker GPT-Red and defender models played against each other, both getting stronger over time. In scenarios where human red-teamers succeeded only 13% of the time, GPT-Red hit 84%. GPT-5.6 was trained against GPT-Red, making it OpenAI’s most robust model against prompt injections (only a 0.05% failure rate). OpenAI just announced something unusual: not a model for users, but an “AI hacker” trained to break its own models. Meet GPT-Red! According to the research publication “Unlocking Self-Improvement for Robustness” released today, GPT-Red automates the work of human security testers (red teams) and does it far better than humans. ...

Update: Claude Mythos 5.1 Is Out (September 1, 2026) The Mythos line was updated with Claude Mythos 5.1 on September 1, 2026. It is the identical model to Claude Fable 5.1, released the same day, with only the level of safeguards differing. It scores 60.9% on Terminal-Bench 4.0 against the Fable build’s 55.8%, and access stays limited to verified organizations under Project Glasswing. For benchmarks, pricing and API changes, read Claude Fable 5.1 . There is a new development every single day in the artificial intelligence world, but this time, the news is truly different. Anthropic announced a brand new model called Claude Mythos Preview on April 7, 2026. Moreover, they brought along a massive cyber defense initiative called Project Glasswing. ...