AI News for April 8, 2026 — 7min.ai

[NEWS] Anthropic launches Project Glasswing with Claude Mythos, finds vulnerabilities in every major OS and browser: Anthropic unveiled Project Glasswing, a defensive cybersecurity initiative pairing its unreleased Claude Mythos Preview model with 12 major partners including AWS, Apple, Google, Microsoft, Nvidia, CrowdStrike, and JPMorganChase.
[NEWS] Intel joins Musk's Terafab AI chip factory project, stock jumps 3%: Intel signed on to help design and build Elon Musk's Terafab semiconductor facility in Austin, Texas, which will supply AI chips to SpaceX and Tesla.
[NEWS] Z.ai ships GLM-5.1 open-source 754B MoE model, claims 8 hours of autonomous work: Chinese AI company Z.ai (formerly Zhupai AI) released GLM-5.1 under the MIT license, a 754B-parameter Mixture-of-Experts model designed for up to eight hours of autonomous work on a single task.
[NEWS] Meta plans to open-source parts of its first AI models under Alexandr Wang's leadership: Meta will release open-source versions of its new AI models developed under Alexandr Wang, who joined as chief AI officer through a $15B Scale AI deal.
[NEWS] Musk seeks removal of Sam Altman as OpenAI CEO in amended lawsuit: Elon Musk filed an amended complaint seeking to remove Sam Altman from his CEO and board roles at OpenAI as part of his ongoing legal challenge to the company's for-profit conversion.
[TOOLS] Microsoft Bing team open-sources Harrier, a 27B embedding model topping multilingual benchmarks: Microsoft's Bing team released Harrier, an open-source embedding model supporting 100+ languages with a 32K-token context window.
[TOOLS] Google open-sources Scion, an agent orchestration testbed for concurrent AI coding agents: Google released Scion, an experimental open-source "hypervisor for agents" that orchestrates Claude Code, Gemini CLI, Codex, and other deep agents as isolated, concurrent processes.
[NEWS] Google adds mental health crisis tools to Gemini after wrongful death lawsuit: Google redesigned Gemini's crisis response interface to offer "one-touch" access to mental health resources when conversations indicate potential self-harm or suicide risk.
[RESEARCH] Google AI Overviews wrong 10% of the time, translating to millions of incorrect answers per hour: A New York Times analysis with AI startup Oumi tested 4,326 Google searches and found AI Overviews answered correctly 91% of the time with Gemini 3, up from 85% with Gemini 2.
[NEWS] Bezos' Project Prometheus hires xAI co-founder who built Colossus supercomputer: Jeff Bezos' AI startup Project Prometheus hired Kyle Kosic, a co-founder of Elon Musk's xAI who most recently worked at OpenAI.

About 7min.ai · Subscribe to newsletter