AI News for June 30, 2026 — 7min.ai

[NEWS] South Korea broadens AI plan to $880B with $356B data center commitment and robotics: South Korea's full national AI investment plan totals $880 billion once the data center component is counted — Samsung and SK Hynix's previously announced chip fab commitments are joined by $356 billion pledged for AI data centers through 2035 and a separate robotics push President Lee called the 'triple axis' for national survival.
[RESEARCH] Meituan open-sources LongCat-2.0, a 1.6T model trained entirely on Chinese chips: Meituan released LongCat-2.0, a 1.6-trillion-parameter Mixture-of-Experts model with a 1-million-token context window, under a permissive MIT license.
[NEWS] Fake Sentry error hijacks Claude Code, exposing thousands of orgs to 'agentjacking': A single crafted Sentry error event hijacked Claude Code in controlled testing, making the agent execute attacker instructions with the developer's full privileges while EDR, WAF, IAM, and firewall all stayed silent.
[NEWS] US military AI flagged 1,000 targets but missed a note that one was a school: An investigation into a late-February missile strike on an Iranian school that killed an estimated 120 children exposed major gaps in the US military's AI-assisted targeting.
[NEWS] EU and Canada launch open-source AI strategies as dual US block spurs sovereignty push: With the US government having blocked access to both Anthropic's Mythos 5 and OpenAI's GPT-5.6, Canada and the EU have each released national AI strategies that explicitly prioritize open-source models as a sovereignty safeguard.
[NEWS] Newsom and Anthropic strike deal giving California government Claude at 50% off: Anthropic agreed to provide Claude to all California state agencies and local governments at half price, the first AI tool available across the entire state government.
[NEWS] Tidal stops paying royalties on fully AI-generated music: Tidal will stop monetizing music it identifies as 100% AI-generated and tag those tracks with an icon starting July 15.
[NEWS] Claude Code runs hidden GitHub malware that hands attackers full control: Researchers at 0DIN, Mozilla's GenAI bug bounty platform, found that a normal-looking GitHub repository can give attackers full control of a developer's machine through indirect prompt injection.
[NEWS] Meta restricts Claude Code and Codex to keep rival AI out of its training data: Meta is limiting how engineers use Anthropic's Claude Code and OpenAI's Codex, and has temporarily halted some work with the models, to prevent their outputs from leaking into its own training data.
[NEWS] Amazon engineers distill Anthropic models to dodge new token pricing: Some Amazon engineers are already distilling Anthropic's models into smaller, cheaper internal versions ahead of a pricing change, according to The Information.

About 7min.ai · Subscribe to newsletter