Google unveiled two new AI models on October 28, 2026. Gemini 4 Carbon targets cybersecurity and software engineering, scoring 88.2% on SEC-bench Pro for vulnerability detection and 71% on CWE-bench for patching. Elon Musk praised the model in a one-word response on X. Alongside it, Google launched Genie 4, which generates interactive virtual worlds with real-time audio for applications in gaming and film planning. Gemini 4 Carbon is rolling out first through Google's Fairwind Program to select cybersecurity defenders, while Genie 4 remains in a research preview for trusted partners.
Canada announced a government equity investment of up to $1.6 billion in Toronto-based AI company Cohere. Prime Minister Mark Carney described the deal as a way to help Cohere build secure AI systems for businesses and governments with tighter control over sensitive data. Cohere committed to creating at least 1,000 new Canadian technical jobs by 2030 and investing at least $1.7 billion in computing infrastructure domestically. Private investors including NVIDIA, Bell, and Schwarz Group joined the commercial round.
Anthropic and OpenAI submitted competing proposals to California Governor Gavin Newsom on how to control dangerous AI models. Anthropic proposed that courts could suspend risky systems, while OpenAI favored requiring strong shutdown controls instead. Both companies agreed on the need for independent safety checks and improved incident reporting. Their submissions followed Newsom's September 2026 executive order on frontier AI safeguards. Former CIA officer Will Hurd, who served on OpenAI's board, warned that bad actors could use AI to attack critical infrastructure like water-treatment plants, saying he was alarmed when seeing OpenAI's GPT-4o model.
Other developments spanned multiple sectors. Salt Lake Valley Emergency Communications Center deployed an AI agent to handle roughly 600,000 non-emergency calls annually, supporting 30 languages and routing information via text or email. Global Fishing Watch partnered with the Allen Institute for AI to track fishing vessels worldwide using satellite data and AI agents, with human oversight guiding final decisions. A new framework called GLIMMER improved blood glucose forecasts for type 1 diabetes, cutting forecasting error by up to 24.6% and achieving 98.4% recall for dangerous glucose events. Mistral launched its Le Chonk model, built for on-premises use by companies looking to reduce reliance on third-party AI. A review published in GeoEnergy Communications found a 180% increase in numerical simulation studies of turbidity currents over five years, with machine learning helping fill data gaps.
Key Takeaways
- <ul><li>Google launched Gemini 4 Carbon for cybersecurity, scoring 88.2% on SEC-bench Pro, and Genie 4 for interactive virtual worlds, both on October 28, 2026.</li><li>Canada invested up to $1.6 billion in Cohere, with NVIDIA, Bell, and Schwarz Group joining as private investors.</li><li>Cohere pledged at least 1,000 new Canadian tech jobs by 2030 and $1.7 billion in domestic computing infrastructure.</li><li>Anthropic and OpenAI submitted different AI safety plans to California Governor Newsom, one favoring court suspensions, the other strong shutdown controls.</li><li>Will Hurd, a former OpenAI board member, warned that AI could be used by bad actors to attack critical infrastructure like water-treatment plants.</li><li>Salt Lake County deployed an AI agent to handle about 600,000 non-emergency calls per year across 30 languages.</li><li>Global Fishing Watch and Ai2 partnered to use AI and satellite data for tracking fishing vessels, announced at Climate Week NYC 2026.</li><li>The GLIMMER framework reduced blood glucose forecasting errors by up to 24.6% for type 1 diabetes patients, with 98.4% recall for dysglycemic events.</li><li>Mistral launched Le Chonk, an on-premises AI model designed for companies to reduce reliance on third-party providers.</li><li>A review in GeoEnergy Communications found a 180% increase in numerical simulation studies of turbidity currents over five years, aided by machine learning.</li></ul>
Elon Musk praises Google's Gemini 4 Carbon AI model
Google released its new Gemini 4 Carbon AI model for cybersecurity and software engineering. Elon Musk called it impressive on X, his one-word response to Google CEO Sundar Pichai's announcement. Google reports the model scored 88.2% on SEC-bench Pro for finding vulnerabilities and 71% on CWE-bench for patching. Gemini 4 Carbon will first be available to select cybersecurity defenders through Google's Fairwind Program before a wider release.
Google launches Gemini 4 Carbon and Genie 4 AI models
Google announced two new AI models on October 28, 2026. Gemini 4 Carbon focuses on cybersecurity, coding and finding software vulnerabilities. Genie 4 creates interactive virtual worlds with real-time audio and can be used for video games, film planning and training. Both models are in limited testing, with Carbon available to select experts and Genie 4 in a research preview for trusted partners.
Canada invests up to $1.6 billion in AI company Cohere
Canada's Prime Minister Mark Carney announced a government equity investment of up to $1.6 billion in Toronto-based AI company Cohere. The investment will help Cohere build secure AI systems for businesses and governments with better control over sensitive data. Cohere plans to create at least 1,000 new Canadian technical jobs by 2030 and invest at least $1.7 billion in computing infrastructure in Canada. Private investors including NVIDIA, Bell and Schwarz Group joined the commercial investment.
AI death prediction tool uses blood analysis to estimate lifespan
Death Clock is an AI-powered tool that predicts when a person will die. The user provides blood samples that are analyzed for biomarkers to create a personalized prediction. A Wired writer tested the tool and found it predicted death in about 10 years but felt the constant reminders of mortality negatively affected mental health. The tool uses machine learning algorithms and data from medical studies, though the developer data sources were not fully transparent.
Will Hurd warns AI threat is real but not zero
Former CIA officer and Texas congressman Will Hurd served on OpenAI's board and worked with CEO Sam Altman. He says he was scared when seeing OpenAI's GPT-4o model, comparing it to a classified TSA briefing in how alarming it was. Hurd warns that bad actors could use AI for cyberattacks on critical infrastructure like water-treatment plants. He believes Russians are already attempting such attacks and stresses the need for regulation as AI technology moves faster than expected.
Salt Lake County adds AI agent to handle non-emergency calls
The Salt Lake Valley Emergency Communications Center will use a new AI agent to answer about 600,000 non-emergency calls each year. The AI will handle routine questions like office hours or how to file a police report. It can send information by text or email and supports 30 languages. The system will not be used for emergency calls.
Global Fishing Watch and Ai2 partner on AI vessel tracking
Global Fishing Watch and Allen Institute for AI (Ai2) announced a partnership to use AI for tracking fishing vessels worldwide. They will use satellite data and AI agents to detect patterns with less human error. The organizations stress that human oversight will still guide final decisions. Their work was announced at Climate Week NYC 2026.
Anthropic and OpenAI differ on stopping risky AI systems
Anthropic and OpenAI submitted different plans to California Governor Gavin Newsom for controlling dangerous AI models. Anthropic wants courts to suspend risky systems, while OpenAI favors requiring strong shutdown controls. Both support independent safety checks and better incident reporting. Their proposals follow Newsom's September 2026 executive order on frontier AI safeguards.
AI and numerical models reshape study of turbidity currents
A new review shows how AI and numerical modeling are changing the study of turbidity currents, which are sediment-laden flows that shape ocean floors and threaten underwater cables. The review found a 180 percent increase in numerical simulation studies over five years. Researchers use models like the one-fluid and two-fluid approaches, with machine learning helping fill gaps. The study was published in GeoEnergy Communications.
New AI framework improves blood glucose forecasts for type 1 diabetes
Researchers developed a training framework called GLIMMER that improves blood glucose predictions for people with type 1 diabetes. It penalizes errors during dangerous glucose swings more heavily than safe-range errors. The approach reduced forecasting error by up to 24.6 percent and achieved 98.4 percent recall for dysglycemic events. It works with multiple model types and was tested on two independent datasets.
Mistral Launches Le Chonk AI Model With Strong Cyber Performance
Mistral announced its new large AI model nicknamed Le Chonk. The model builds on a public preview released earlier in October 2026. It shows strong performance on cybersecurity tasks. The model is designed for companies to use on their own systems. This reduces reliance on third-party AI models.
Sources
- Elon Musk's One-Word Response to Sundar Pichai's Post On New AI Model
- Google launches Gemini 4 Carbon and Genie 4 AI models: Key features, uses and availability
- Canada announces landmark investment in Cohere to scale secure, sovereign AI
- I Gave My Blood to the AI Death Clock
- Will Hurd Worked With Sam Altman. His (P)Doom Is “Not Zero.”
- AI agent coming to Salt Lake County non-emergency line
- Global Fishing Watch and Ai2 announce another partnership to use AI in vessel tracking globally
- Anthropic and OpenAI diverge on who should pull the plug on risky AI
- Turbidity Currents Meet AI: Review Maps Numerical Models and Machine Learning
- Lightweight AI training framework sharpens dangerous blood glucose forecasts in type 1 diabetes
- Mistral Touts Cyber Capabilities of Its ‘Chonk’ AI Model
Comments
Please log in to post a comment.