AI News for April 15, 2026 — 7min.ai

[NEWS] Amazon buys Globalstar for $11.6B, becomes primary satellite provider for iPhone and Apple Watch: Amazon agreed to acquire satellite operator Globalstar for $11.57 billion and simultaneously signed a deal with Apple to provide satellite connectivity for current and future iPhone and Apple Watch features.
[NEWS] Anthropic flooded with investor offers at $800B valuation, more than double its last round: VCs are offering to invest in Anthropic at valuations as high as $800B, more than double its $380B post-money from February's Series G.
[NEWS] OpenAI's $852B valuation faces investor skepticism as Anthropic surges: Some OpenAI investors are questioning the company's $852B valuation as Anthropic's revenue surges and coding tools gain traction.
[NEWS] Nvidia launches Ising, first open-source AI models for quantum computing: Nvidia unveiled Ising, an open-source family of AI models designed to solve quantum computing's core processor calibration and error-correction problems.
[RESEARCH] UK AISI: Mythos is first AI to autonomously compromise enterprise networks end-to-end: The UK AI Security Institute published independent evaluation of Anthropic's Mythos Preview showing 73% success on expert-level capture-the-flag challenges and 85%+ on apprentice-level tasks.
[NEWS] GPT-5.4-Cyber launches to limited users, confirming OpenAI's Mythos rival: OpenAI officially launched GPT-5.4-Cyber to a limited group of vetted users, confirming last week's reports of a restricted cybersecurity model in development.
[TOOLS] Claude Code launches Routines for automated bug fixes, code review, and deployments: Anthropic introduced Routines for Claude Code: automated processes that run independently on schedules, API calls, or GitHub events without requiring a user's local machine.
[TOOLS] Google launches Skills in Chrome, turning Gemini prompts into one-click workflows: Google shipped Skills in Chrome, letting users save frequently used Gemini prompts and rerun them across any webpage with a single click.
[RESEARCH] Stanford AI Index: SWE-bench hit 100% in one year, US-China performance gap 'practically closed': The Stanford AI Index 2026 documents dramatic benchmark progress beyond yesterday's public-trust findings: AI models now outperform humans on PhD-level science questions, and the coding benchmark SWE-bench jumped from 60% to near-100% in a single year.
[NEWS] Fortune: Anthropic quietly reduced Claude's effort defaults to cut token costs ahead of IPO: Fortune's investigation identified the specific mechanism behind widespread Claude performance complaints: Anthropic quietly reduced the model's default "effort" level to economize on tokens.

About 7min.ai · Subscribe to newsletter