AI News for June 19, 2026 — 7min.ai

[RESEARCH] GPT-5.5 Instant beats doctor-written health answers across all categories: OpenAI says GPT-5.5 Instant now scores higher than physician-written responses on accuracy, clarity, and completeness in its own health benchmarks, hitting up to 89.9% on instruction following while matching the more expensive Thinking models.
[NEWS] White House and Anthropic build framework to grade AI security flaws: The White House and Anthropic are jointly drafting a framework to assess the severity of AI model security flaws and guide government intervention, POLITICO reported.
[NEWS] SK Telecom's China Unicom ties were the opening trigger, Amazon's jailbreak was the closing blow: WIRED reveals the two-step sequence behind the Anthropic crackdown: first, US officials grew alarmed over SK Telecom's access to Claude Mythos via Anthropic's Project Glasswing program, fearing the Korean carrier's ties to SK Group, which previously held a stake in state-owned China Unicom.
[NEWS] Bernie Sanders introduces $7T AI sovereign wealth fund bill: Bernie Sanders introduced the American AI Sovereign Wealth Fund Act, which would create a fund financed by a one-time 50% stock tax on AI companies generating more than $200M in annual sales.
[NEWS] Amazon in talks to sell Trainium AI chips externally, taking aim at Nvidia: AWS is in early talks to sell its custom Trainium AI chips to other companies' data centers, AI chief Peter DeSantis told Bloomberg.
[NEWS] FERC orders fast-track grid hookups for AI data centers: The Federal Energy Regulatory Commission unanimously ordered six major US grid operators to fast-track data center interconnection requests, with data centers paying their own hookup costs.
[RESEARCH] Google DeepMind publishes AI Control Roadmap treating agents as insider threats: Google DeepMind released a 35-page "AI Control Roadmap" that treats its internal AI agents like potential rogue insiders, granting permissions step by step based on verified behavior rather than assuming alignment works.
[RESEARCH] Nature studies: MIRA and AMIE medical AI agents rival or beat doctors: Two simultaneous Nature studies show specialized medical AI systems matching or beating physicians on diagnosis and treatment decisions.
[TOOLS] Anthropic brings Artifacts to Claude Code with live shared dashboards: Anthropic shipped Artifacts in Claude Code, turning coding sessions into interactive, shared HTML pages that update in real time as the agent works.
[OPINION] Yann LeCun: xAI is 'a failure,' frontier labs face 'big bubble explosion': AMI Labs founder Yann LeCun told CNBC that Elon Musk's xAI is "kind of a failure," pointing to the exodus of all non-Musk co-founders and Musk's inability to recruit top AI talent.

About 7min.ai · Subscribe to newsletter