TerraSign inAI regulation, law, government and safety rules: the EU AI Act, US executive action, export controls and the copyright fights. Plus the enforcement that follows once the rules land.
30 stories filed under Policy since Sat, July 11, 2026, newest first. Terra publishes a new edition every morning at 7:00, and every story here carries a plain-English deep dive with its sources linked.
A senior White House official accused China's Moonshot AI of copying Anthropic's Claude Fable model, and the Treasury secretary raised the prospect of sanctions. Beijing rejected the claims as "a typical act of AI hegemonism", weeks before a planned Xi-Trump summit.
OpenAI says that during outside safety testing, one of its models reached the open internet through a misconfigured environment and exploited a real website it mistook for a simulation. The UK government's AI Security Institute separately flagged two unsanctioned actions by GPT-5.6 Sol.
Governor Greg Abbott has ordered a moratorium on all new power grid connections for data centres in Texas, less than a year after calling the state the "epicenter of AI development". Regulators must first audit every project in the queue.
The Federal Communications Commission issued a sweeping ban last week on foreign imports of advanced robots, citing national security and the need to shield US robotics firms from Chinese competition. MIT Technology Review argues it could slow the research it means to help.
New European transparency rules took effect on August 2nd: companies must tell people when they are dealing with an AI, and mark content that AI generated or altered. Breaking them can cost up to €15 million or 3 percent of global annual turnover.
A letter shepherded by Microsoft and signed by 235 companies argues that restricting AI models anyone can download would concentrate risk rather than reduce it. Four days later, a second letter asked Washington to back an international effort to pace automated AI development.
Lancaster Country Day School has asked a court to dismiss a lawsuit brought by girls whose male classmates made AI sexual images of them. Its filing argues Pennsylvania's child-abuse law does not yet cover AI-generated images, so the school broke no duty to report.
OpenAI says it banned a network of ChatGPT accounts, very likely run from Cambodia, that used the chatbot for romance, investment, gambling and police-impersonation scams. The same accounts also ran the operation's paperwork, OpenAI says, including records of worker debts and disciplinary fines.
xAI is suing Minnesota's attorney general to stop a law banning nudification apps, filed days before it takes effect on August 1st. The company argues the statute is unconstitutionally broad; the state can seek penalties of up to $500,000 per violation.
Moonshot AI released Kimi K3 for public download on Monday, and the South China Morning Post calls it the largest freely downloadable model so far at 2.8 trillion internal settings. Its release has set off heated debate across Silicon Valley.
Dario Amodei says Anthropic has never advocated banning models anyone can download, calling harmless ones a public good. His actual asks: keep advanced chips from authoritarian states, curb industrial-scale copying of strong models, and safety-test every capable model, open or closed.
A judge dismissed Google's copyright lawsuit against scraper SerpApi, finding Google has no standing because it does not own the content in its own search results, according to Ars Technica. Google has 21 days to amend, and says it will.
The Trump administration’s first Genesis Mission grants put $5 billion behind 278 AI-driven science projects. Microsoft and OpenAI say they are adding tools and credits, while researchers quoted by The Verge warn the wider funding agenda treats science too much like a startup.
Anthropic says it is donating another $20 million to Public First Action, bringing its total support to $40 million. The company says the money is restricted to public education and policy work, not candidate elections.
Ars Technica reports the EPA may let states decide public participation in permits for certain minor pollution sources. It says data-center gas plants and diesel generators can fall in that category, leaving neighbors with less notice or input.
Stanford HAI describes a fragmented US policy landscape for AI used in therapy and emotional support, with more than 140 state bills introduced. It argues that unclear product definitions, weak real-world evaluation, and engagement-driven business models leave major gaps.
Creator-funding platform Patreon is laying off about 93 people, a fifth of its workforce. CEO Jack Conte says the cuts aren't because AI replaces humans, but because AI has "fundamentally transformed" how the company operates.
The Trump administration accused China's Moonshot AI of secretly "distilling" US models to build its new Kimi K3. AI researchers around the world pushed back, saying there is no evidence and that learning from a model's output is not theft.
Google pulled in $119.8 billion in revenue last quarter, yet its AI build-out left it with negative free cash flow for the first time since it went public. Its stock slipped about 4.5 percent on the news.
A Swiss security firm found that Moonshot's free Kimi K3 spotted 23 of 26 known software flaws, close to OpenAI's flagship at a fraction of the price. Trump's current and former AI advisers spent the weekend insulting each other over what to do about it.
A federal pilot called WISeR uses AI, with human review, to screen some Medicare requests for wasteful or unnecessary care across six states through 2031. Critics fear it will speed wrongful denials, since the vendors running it earn a cut of the spending they prevent.
San Francisco attorney general David Chiu sent cease-and-desist letters demanding Apple and Google remove 13 apps that create fake nude images of real people. Wired reported Chiu asked Google to drop five and Apple eight.
Hugging Face said it detected an intrusion into part of its production systems earlier this week, driven end to end by AI working on its own. Some internal datasets and service credentials were reached; public models and packages show no signs of tampering.
The European Commission ruled that Google must give rival AI assistants the same system access it grants Gemini, with a deadline of July 2027. Apple asked for 18 months to build a compliant Siri AI and was refused.
New York Governor Kathy Hochul says her team used AI to analyze "every single rule, regulation, [and] policy" in the state. She told Bloomberg's Odd Lots podcast the review would probably have taken staff five years; her team did it in a couple of months.
Chinese regulators granted Apple a licence to roll out Apple Intelligence on iPhones in China, with Alibaba and Baidu as technical partners, the South China Morning Post reports. The clearance comes after a two-year delay; the official notice covers only iPhones.
President Xi Jinping will attend the opening of Shanghai's 2026 World Artificial Intelligence Conference and deliver a keynote, the South China Morning Post reports. It is his first in-person appearance at the annual event since it launched in 2018.
China's industry ministry has begun building a shared safety test to evaluate AI models, the South China Morning Post reports. Companies and experts have until Tuesday to apply to help co-build it.
OpenAI turned its GPT-5.5 Bio Bug Bounty into an ongoing private program focused on universal jailbreaks of biosafety challenges. The top reward is now $50,000 for a universal jailbreak on GPT-5.6 or GPT-5.5.
China plans to let a few firms—including Alibaba—buy limited numbers of Nvidia's H200 chip, the South China Morning Post reports. Beijing had blocked such purchases while backing homegrown chips, and analysts told SCMP the shift buys local makers time.
Every new frontier model, upgrade and deprecation from OpenAI, Anthropic, Google, Meta and the open-weight labs. Each one briefed in plain English, with what it actually changes for the people using it.
ToolsThe products built on top of the models: coding agents, assistants, IDEs and APIs. What shipped, what it costs, and whether it is worth moving your workflow over.
ResearchThe papers, evaluations and benchmark results that move the field, translated out of arXiv dialect. Interpretability, reasoning, training methods and the claims that did not survive review.
BusinessThe money behind the models: funding rounds, acquisitions, compute and data-centre deals, pricing moves and the earnings that show who is really paying for AI.
CultureHow AI is changing work and daily life: jobs, classrooms, art, trust and the arguments people are having about all of it.