AI News for April 22, 2026 — 7min.ai

[TOOLS] OpenAI launches ChatGPT Images 2.0 with reasoning, infographics, and flawless text: OpenAI released ChatGPT Images 2.0, a new GPT-Image-2 model with "thinking" capabilities that can search the web, reason about image structure before generating, and produce up to eight images at once with consistent characters.
[NEWS] SpaceX strikes $60B option to buy Cursor, $10B backup for partnership: SpaceX announced a deal giving it the right to acquire AI coding startup Cursor for $60B later this year, or pay $10B for its work if no acquisition happens.
[NEWS] Unauthorized group gains access to Anthropic's Mythos via third-party vendor: A private online forum obtained access to Mythos, the cybersecurity model Anthropic said was too dangerous for broad release, through a third-party vendor environment.
[NEWS] Meta starts capturing employee keystrokes and mouse movements to train AI: Meta launched internal software that records US employees' keystrokes, mouse movements, and periodic screenshots to build training data for its future AI agents.
[NEWS] Florida opens criminal probe of OpenAI over ChatGPT's role in mass shooting: Florida Attorney General James Uthmeier subpoenaed OpenAI and launched a criminal investigation into whether ChatGPT aided or abetted the Florida State University mass shooting that killed two and wounded six.
[RESEARCH] Mozilla: Mythos found 271 zero-day vulnerabilities in Firefox 150: Mozilla revealed that early access to Anthropic's Mythos Preview helped identify 271 security vulnerabilities in this week's Firefox 150 release.
[TOOLS] Google launches Deep Research and Deep Research Max agents on Gemini 3.1 Pro: Google shipped Deep Research and Deep Research Max, two autonomous research agents built on Gemini 3.1 Pro that let developers fuse open web data with proprietary enterprise data through a single API call.
[NEWS] Core Automation launches, poaching top researchers from Anthropic and DeepMind: A new AI startup founded by ex-OpenAI VP Jerry Tworek emerged as "the world's most automated AI lab," having recruited Anthropic researcher Rohan Anil and DeepMind Gemini researcher Anmol Gulati, among others.
[RESEARCH] Three AI coding agents leak secrets via single prompt injection attack: Security researchers demonstrated that a single malicious instruction in a GitHub pull request title caused Anthropic's Claude Code Security Review, Google's Gemini CLI Action, and GitHub's Copilot Agent to post their own API keys as comments.
[NEWS] Sam Altman calls Anthropic's Mythos "fear-based marketing": On the Core Memory podcast, OpenAI CEO Sam Altman publicly attacked Anthropic's handling of Mythos, framing the company's warnings as a marketing tactic to justify keeping AI in elite hands.

About 7min.ai · Subscribe to newsletter