AI News for May 24, 2026 — 7min.ai

[NEWS] Anthropic Mythos finds 10,000 critical vulnerabilities in a month, outpaces patching: Anthropic's Claude Mythos Preview, run by about 50 Project Glasswing partners, surfaced more than 10,000 high- or critical-severity flaws in system-critical software in its first month.
[NEWS] DeepSeek makes 75% V4-Pro discount permanent, undercuts GPT-5.5 by 34x on output: DeepSeek converted its temporary 75% cut on flagship V4-Pro into permanent pricing: $0.435 per million input tokens and $0.87 per million output tokens.
[OPINION] Microsoft finds AI agents cost more than the human employees they replace: Internal Microsoft analysis cited by Fortune shows that running AI agents on enterprise tasks is currently more expensive than paying the human workers doing the same jobs, once token consumption is fully accounted for.
[TOOLS] Claude Code has won the AI coding wars inside startups, BI survey finds: A Business Insider survey of two dozen founders and VCs found Anthropic's Claude Code is now the default AI coding tool inside startups, praised for handling complex engineering tasks and autonomous workflows.
[OPINION] Pichai reframes links as just "part" of Google Search: Asked in a post-I/O podcast whether Google would keep showing links, CEO Sundar Pichai said "sources and links will always be there as part of it." The word "part" marks a notable downgrade from the open web's historical status as Search's foundation.
[NEWS] UC Berkeley Law bans AI from nearly all graded work starting summer 2026: UC Berkeley School of Law will prohibit students from using AI to brainstorm, draft, outline, write, revise, translate, or proofread anything they submit for grades, with exams completely off-limits.
[NEWS] OpenAI posts $445K safety role to prepare for recursive self-improvement: OpenAI's Preparedness team is hiring a researcher at $295K to $445K to "support preparations for recursive self-improvement," tracking progress toward automating its own technical staff.
[NEWS] Nvidia CEO Huang urges Super Micro to tighten compliance after Taiwan smuggling arrests: Jensen Huang publicly pressured Super Micro Computer to tighten its compliance practices after Taiwanese prosecutors detained three people this week for allegedly making fraudulent export declarations on AI servers built by Nvidia's US partner.
[NEWS] Hyperscaler AI debt binge sparks credit-derivatives bonanza on Wall Street: As big tech companies raise hundreds of billions to fund AI buildouts, Wall Street banks say they must trade ever more credit derivatives to keep doing business with the hyperscalers.
[OPINION] Former NASA robotics chief warns US is building the wrong humanoids: Ex-NASA robotics chief Robert Ambrose argues that US humanoid programs are optimizing for sprint demos rather than deployment marathons.

About 7min.ai · Subscribe to newsletter