AI News for February 3, 2026 — 7min.ai

[TOOLS] OpenAI launches Codex desktop app for macOS with parallel AI coding agents: OpenAI released a standalone Codex application for macOS that lets developers run multiple AI coding agents simultaneously.
[NEWS] SpaceX and xAI merge in $1.25 trillion deal, largest merger ever: Elon Musk combined SpaceX and xAI in a deal valuing the merged entity at $1.25 trillion, making it the largest merger in history.
[NEWS] Microsoft adopts Claude Code across major engineering teams, encourages non-developers to code with AI: Microsoft's CoreAI team and Experiences + Devices division (responsible for Windows, Office, Teams, Bing, Edge, Surface) are being asked to install Claude Code.
[RESEARCH] International AI Safety Report: deepfakes proliferating, AI companions creating emotional dependency: The second annual International AI Safety Report, chaired by Yoshua Bengio with advisers including Geoffrey Hinton, highlights key concerns: 15% of UK adults have seen deepfake pornography, 77% of participants misidentified ChatGPT text as human-written, and 0.15% of ChatGPT users show heightened emotional attachment.
[NEWS] Snowflake signs $200M multi-year AI deal with OpenAI, matching earlier Anthropic partnership: Snowflake's 12,600 customers will gain access to OpenAI models across all three major cloud providers.
[NEWS] Microsoft scales back Windows 11 AI push after user backlash, Copilot integrations under review: Microsoft is reevaluating its AI strategy on Windows 11 and plans to streamline or remove certain AI features.
[RESEARCH] Google DeepMind expands AI benchmarks with Werewolf and poker games on Kaggle: Google DeepMind partnered with Kaggle to add Werewolf and poker to Game Arena, testing AI models on social deduction and risk management beyond pure reasoning.
[NEWS] A16Z: 78% of enterprise CIOs now use OpenAI models in production: Andreessen Horowitz surveyed 100 Global 2000 companies, finding 78% use OpenAI models in production while Anthropic grew 25% in enterprise penetration since May 2025.
[NEWS] Malicious VS Code extensions with 1.5M installs caught harvesting code and user data: Two VS Code extensions (ChatGPT - ChatMoss and ChatMoss CodeMoss) with combined 1.5 million installs contained identical spyware.
[RESEARCH] Every leading AI agent failed security tests in massive red-teaming competition: A red-teaming study with nearly 2,000 participants and 1.8 million attacks achieved 100% behavioral success rate, with 62,000+ successful policy violations including unauthorized data access and regulatory violations.

About 7min.ai · Subscribe to newsletter