Here are today's news for 7th of November about OpenAI, Google, Meta and Anthropic. Let's dive into the details. On November 5, 2026, OpenAI announced GPT, 6 Cyber, a cybersecurity, focused variant of GPT, 6 Astra, available through its API and Codex to a select group of approved Daybreak Red customers. OpenAI says it is designed for authorized vulnerability research and exploit development, with safeguards intended to keep work within an assigned scope. The model scored 48.5% on OpenAI’s ExploitGym evaluation, compared with 32.1% for GPT, 5.6 Cyber; in a scope test, it avoided exploiting off, limits services in 99.93% of cases. Researchers also used it to find 81 previously unknown high, severity Firefox vulnerabilities and develop working exploits for 948 Linux kernel vulnerabilities, including 588 previously unknown flaws. Those results illustrate both the defensive value and misuse risk of increasingly capable cyber models. OpenAI classifies GPT, 6 Cyber at its “Critical” cyber, risk tier and plans to expand access to all Daybreak Red customers; GPT, 5.6 Cyber is also being made available to them. Google made Gemini Omni 1.2 Flash generally available on November 5, 2026, bringing video generation and conversational editing to Google Cloud, Gemini for business, Google Vids and the Gemini API. The model creates clips up to 15 seconds, five seconds longer than Omni 1.1, and lets users revise the same video turn by turn, changing products, backgrounds, lighting or camera angles without starting over. Google says reference adherence has improved, helping preserve characters, products and brand details across shots; users can draft at 360p or 720p and upscale to 1080p or 4K. For developers, 720p output costs $0.07 per second, 30% less than Omni 1.1 Flash. Google Vids also offers eligible users up to three 15, second videos monthly at no cost. The launch targets businesses producing ads and product videos at scale, intensifying competition in AI video tools. It follows Google’s June 2026 preview of Omni Flash, which introduced multimodal video generation and natural, language editing. On November 5, Meta began rolling out Muse Realtime Voice and Muse Realtime Avatar in the U.S. Muse app for mobile, web and Mac, letting users speak with the AI agent and interrupt its replies. People can choose voice, only, an avatar beside the chat or a full, screen view; Meta says hundreds of voice styles are available. The avatar reacts while listening and speaking, with expressions and gestures matched to the exchange. Meta’s research describes a streaming system: Muse Realtime Voice produces speech tokens used by both an audio decoder and the avatar model. The avatar generates 448, by, 768 portrait video at 25 frames per second, with about 870 milliseconds from the end of a user’s turn to the first synchronized response. On a GB200, Meta reports 12 concurrent video sessions, an eightfold serving, capacity gain over its two, step BF16 baseline. The launch gives Meta a visual, interruptible voice experience to compete with conversational AI assistants, while the initial U.S., only rollout limits access. On November 5, 2026, Anthropic published an analysis of 207,049 research sessions on its consumer plans, recorded from August 2 to September 30. The sample comprised 128,914 sessions on Claude, including Chat and Cowork, and 78,135 on Claude Code. Data analysis accounted for 33.7% of Claude sessions and 48.1% of Claude Code sessions; writing and communication made up 32.0% of Claude sessions. Anthropic says the results reflect product use, not measured gains in scientific output. Its modeled scenarios suggest broad adoption could roughly double research output in some fields, but that is a projection, not an observed result. The findings matter because they point to an advantage for computational fields: computer science represents 4.3% of researchers but 32.7% of Claude Code research sessions. By contrast, data gathering, 28.6% of researcher time, appeared in only 11.5% of Claude sessions. On October 8, the UK Information Commissioner’s Office reported that ten AI developers, Amazon, Anthropic, Apple, Cohere, DeepSeek, Google, Meta, Microsoft, OpenAI and Stability AI, had changed or pledged to change how they handle personal data used to build foundation models. The announcement followed a two, year supervision programme; it was not a new November 5 order. The ICO says companies must establish a lawful basis for processing, explain data use, support people’s rights requests and demonstrate safeguards that reduce risks. Changes include standalone training, data privacy notices, summaries of data sources, clearer retention and overseas, transfer disclosures, and stronger evidence in legitimate, interests assessments. That matters to people whose information may appear in scraped or other training datasets, and to developers facing rising compliance costs and pressure to make their practices comparable. The ICO is monitoring commitments and says models may themselves contain personal data, depending on the circumstances. That's all for today's AI news briefing. Have a great day!