AI Industry Daily News

Trump Administration Orders Anthropic to Block Foreign Access to Claude Fable 5 and Mythos 5

The Trump administration on June 12, 2026 instructed Anthropic to prevent foreign nationals from using its most advanced models, Claude Fable 5 and Claude Mythos 5, under export-control authority. The directive from Commerce Secretary Howard Lutnick covers all foreign nationals inside and outside the US, including Anthropic's own foreign-national employees.

Because selectively restricting access by nationality is technically difficult for AI models, Anthropic said it had to temporarily disable both models for all customers to comply; other Claude models are unaffected. The two models had been released only days earlier, around June 9. As of June 15, Anthropic's technical staff were meeting with Commerce Department officials to negotiate lifting the restriction.

The directive is tied to a limited jailbreak concern around Fable 5's safety guardrails that the government judged a cybersecurity threat, though details remain undisclosed. Anthropic noted similar capabilities exist in other public models such as GPT-5.5. Critics describe the move as a de facto licensing regime for frontier AI models, a marked shift from the administration's June 2 executive order that had favored voluntary frameworks.

Sanders Plans Bill to Transfer 50% of Major AI Companies' Equity to a Federal Fund

Senator Bernie Sanders said in June 2026 he will introduce the American AI Sovereign Wealth Fund Act, which would impose a one-time tax paid in equity—half of the shares of major AI companies including OpenAI, Anthropic and xAI—transferred to a federal national sovereign wealth fund.

Under the plan, outlined in a New York Times guest essay and a statement, the government would acquire voting shares and equivalent board representation, allowing it to block corporate decisions deemed harmful to citizens. Proceeds would fund direct dividends to US citizens and public programs in health, education and housing. Sanders cited Norway's oil-funded sovereign wealth fund and Alaska's Permanent Fund Dividend as precedents.

The full bill text is not yet public; estimates referenced an AI revenue threshold of $200 million per year for covered companies and a fund of roughly $7 trillion. Sanders argued generative AI was built on humanity's collective knowledge used without permission or compensation.

Justice Department Intervenes on xAI's Side in Mississippi Data Center Lawsuit

The US Justice Department has intervened in a lawsuit concerning xAI's data center in Mississippi, taking a position in support of xAI. The case puts the siting of AI infrastructure and the relationship between state and federal authorities at the center of a political dispute.

The intervention adds to a broader pattern of federal involvement in AI infrastructure and policy decisions.

OpenAI Releases LifeSciBench and Reports GPT-5.4 Tested in Drug Discovery

OpenAI announced LifeSciBench, a benchmark for life-sciences research built from 750 tasks and supervised by 173 scientists. The release positions the work as an effort to evaluate AI on real scientific tasks.

Alongside the benchmark, OpenAI reported that GPT-5.4 proposed improvements to a novel reaction in a drug-discovery medicinal chemistry project, and that the suggestion was verified experimentally. The company framed the result as AI for Science moving into a demonstration phase.

Other groups are advancing related work, including Tongyi Lab's integrated AI-for-Science model LOGOS, referenced in the source as announcements only.