This brief covers the trailing ~72 hours (November 5 – November 7, 2026). Items are drawn from the originating organizations’ own channels and date-checked; it was a moderate window led by OpenAI’s cybersecurity model and new video and open-weights-bound models from Google and Thinking Machines.
OpenAI introduces GPT-6 Cyber, a model for finding and fixing exploitable vulnerabilities
OpenAI · November 5, 2026
OpenAI announced GPT-6 Cyber, a cybersecurity-focused model built to find exploitable vulnerabilities and develop fixes, framed around “aligned” cybersecurity agents.
“Introducing GPT-6 Cyber, OpenAI’s cybersecurity model for finding exploitable vulnerabilities and developing fixes…” — OpenAI
Source: GPT-6 Cyber: Advancing the frontier of aligned cybersecurity agents
Google releases Gemini Omni 1.2 Flash and brings it to Google Vids
Google · November 5, 2026
Google launched Gemini Omni 1.2 Flash, a fast, cost-effective video model with longer clips. In Google Vids, anyone with a Google account or Workspace subscription can make three 15-second AI videos per month at no cost, with more consistent characters and scenes from reference images and more natural speech narration.
“If you can describe it, you can make it.” — Google
Sources: Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips; Generate longer AI videos with Omni 1.2 Flash in Google Vids
Thinking Machines Lab introduces Inkling 2 Small
Thinking Machines Lab · November 5, 2026
Thinking Machines released Inkling 2 Small, a mixture-of-experts model with 594B total parameters (17B active) that accepts image inputs. It is initially available through Tinker with closed weights; the company plans to release full weights under Apache 2.0 after external safety testing.
“Today we are introducing our second model generation with Inkling 2 Small.” — Thinking Machines Lab
Source: Inkling 2 Small
OpenAI introduces BioBasicsBench for AI agents in wet-lab science
OpenAI · November 6, 2026
OpenAI introduced BioBasicsBench, a 100-task benchmark measuring how well AI agents handle everyday wet-lab work such as interpreting raw experimental data.
“OpenAI introduces BioBasicsBench, a 100-task benchmark measuring how well AI agents interpret raw experimental data…” — OpenAI
Source: Measuring how AI agents handle everyday wet-lab science
OpenAI partners with Hacktron to find vulnerabilities in its own systems
OpenAI · November 6, 2026
OpenAI announced a partnership with security firm Hacktron to probe its systems for vulnerabilities, including connections with third-party services, continuing the security push described in the previous brief.
“OpenAI is partnering with Hacktron to find vulnerabilities, test connections with third-party services…” — OpenAI
Source: Partnering with Hacktron to find vulnerabilities in OpenAI’s systems
This brief covers the trailing ~72 hours (November 5 – November 7, 2026).
Primary sources: