In partnership with
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
Today’s Highlights:
🎙️ OpenAI Unveils 3 Audio Models for Real-Time Voice Tasks
🧠 Google DeepMind Unveils AI Co-Mathematician
🛡️ OpenAI Adds ‘Trusted Contact’ Safety Feature
💸 Anthropic Reportedly Commits $200B to Google Cloud Infrastructure
🖼️ Gemini File Search Now Supports Image Understanding
💼 Anthropic Pushes Claude Into Wall Street Workflows
🤖 Microsoft Expands Copilot Into Full Task Automation Tool
🛡️ U.S. Government Gets Early Access to Frontier AI Models
🚀 Grammarly Rebrands to Superhuman, Drops AI Assistant
Read Time: 4 mintue
LATEST AI NEWS
OPENAI
Image Source: OpenAI
OpenAI has launched a new set of real-time voice models designed to make AI conversations feel more natural, intelligent, and action-oriented.
The Details:
OpenAI introduced GPT-Realtime-2, a voice model built with GPT-5-class reasoning for more complex conversations
New model can listen, respond, reason, and take actions in real time instead of simple voice replies
OpenAI also launched GPT-Realtime-Translate, supporting 70+ input languages and 13 output languages for live translation
New GPT-Realtime-Whisper provides real-time speech-to-text transcription as conversations happen
All models are available through OpenAI’s Realtime API for developers building voice-based apps and services
OpenAI says the tools are designed for customer support, education, creator platforms, events, and enterprise workflows
Built-in guardrails aim to prevent misuse like spam, scams, or harmful impersonation attempts
GPT-Realtime-2 is billed by tokens, while Translate and Whisper are billed per minute of usage
Why it matters:
Voice AI is moving beyond assistants that simply answer questions. OpenAI is building systems that can actively participate in conversations, understand nuance, and complete tasks in real time, bringing AI closer to human-like interaction.
Google DeepMind has introduced an AI co-mathematician designed to collaborate with researchers on unsolved math problems, pushing AI beyond assistants into real scientific discovery.
The Details:
Built on Gemini 3.1, the system uses multiple AI agents working together like a research team
Designed for open-ended math workflows including ideation, theorem proving, literature search, and theory building
Achieved a new high score on FrontierMath Tier 4, solving nearly half of one of the toughest AI math benchmarks
One Oxford professor reportedly used the system to help solve a long-standing open math problem
Includes a verifier system that checks proofs, identifies flaws, and iteratively improves solutions
DeepMind says the system can admit failure and restart reasoning paths instead of confidently hallucinating answers
Part of Google’s broader push toward AI systems that accelerate scientific and mathematical research
Why it matters:
This moves AI closer to becoming a true research collaborator. Instead of just answering questions, these systems are beginning to generate ideas, test hypotheses, and contribute to frontier scientific discovery.
OPENAI
Image Source: Shutterstock
OpenAI has launched a new ChatGPT safeguard called Trusted Contact, designed to alert someone close to a user if conversations suggest possible self-harm.
The Details:
Adult users can now add a trusted contact like a friend or family member to their ChatGPT account
If conversations show signs of possible self-harm, ChatGPT encourages the user to reach out to that person
OpenAI may also send the trusted contact a brief alert via email, text, or app notification
Notifications avoid sharing detailed chat content to protect user privacy
System combines automated detection with human review, with OpenAI aiming to review serious cases within one hour
Feature follows earlier parental safety tools introduced for teen accounts
Trusted Contact is optional, and users can still create multiple ChatGPT accounts outside the system
OpenAI says the move is part of broader efforts to make AI safer during emotional or mental health crises
Why it matters:
AI systems are becoming deeply personal, which means companies are now being forced to handle emotional safety, not just technical performance. This feature shows how AI platforms are evolving into systems that actively intervene during high-risk situations.
AI TUTORIAL
Generate complex, high-functioning websites from a single prompt. Unlike other builders, Mironik gives you a built-in visual editor to fine-tune every detail without writing any code.
Steps to Follow:
Open Mironik.com and enter a brief description of the website or landing page you want to create.
Let the AI generate the entire site structure, including interactive sections, carousels, and nested components.
Use the Visual Editor to click and edit any element, adjust layouts, or change styles with pixel-level control.
Add smart forms, accordions, or tabs by simply prompting the AI to "add a section for X."
Hit Publish to take your site live with built-in hosting, SSL security, and a global CDN.
✨ Stop fighting with rigid templates: Mironik turns your ideas into fully editable, professional websites 10x faster.
MORE AI & TECH NEWS
Anthropic is marketing Claude to banks, insurers, and fintech firms for tasks like pitch books, compliance reviews, and fund reports. The system also includes source tracing, helping finance teams verify numbers and audit AI-generated work.
Google upgraded Gemma 4 with multi-token prediction, allowing the model to generate several words simultaneously instead of one at a time. The change boosts speed by up to 3x without sacrificing quality, improving chat, coding, and mobile AI performance.
Anthropic is reportedly planning to spend $200 billion over five years on Google Cloud and AI chips. The deal highlights the enormous computing costs required to train and run frontier AI systems at scale.
Google expanded Gemini API File Search to analyze both text and images. The update also adds custom labels and page citations, helping developers build more accurate search tools with traceable AI-generated answers.
Microsoft upgraded Copilot Cowork to handle inbox management, research, documents, and web workflows. The system now supports reusable skills, mobile apps, and integrations with enterprise platforms like Power BI, Dynamics 365, and monday.com.
Google, Microsoft, and xAI agreed to let U.S. officials test advanced AI systems before public release. The initiative focuses on identifying cybersecurity and national security risks tied to powerful frontier models.
AI chipmaker Cerebras is seeking a valuation of up to $26.6 billion in its upcoming IPO. The offering will test investor appetite for AI infrastructure companies competing directly against Nvidia.
NEW AI TOOLS TO 10X YOUR PRODUCTIVITY
⚡ NoDu AI – Automate documents, PDFs, and content workflows instantly using AI.
🎬 cutback.video – Turn long videos into highlight reels with smart AI clip selection.
🎥 2.video – Convert text, images, or ideas into polished videos in seconds.
🎵 anymelo.ai – Generate, remix, or edit music and audio with simple prompts.
🧠 Mistral AI – Lightning-fast AI models for text, reasoning, and data tasks.
📚 Vidsbo – Turn your ideas or scripts into clear, visual storyboards automatically.
🔎 ForumScout – Track brand mentions across forums and find high-intent leads.
AI MEME OF THE DAY
SPONSOR US
AI Insights newsletter is read by thousands of AI and Tech professionals/enthusiasts around the world.
Get in touch to get your product seen today!
Or email us at: [email protected]
THANK YOU FOR READING
FEEDBACK
How would you rate today's newsletter?
We'd love to hear your feedback or any interesting thoughts you have!
Please share by replying to this email.
