Skip to content

#政策/监管

4 today

Jul 21Tuesday

New York Times Chinese

US treats AI like nukes, China treats it like nuclear energy

Ross Douthat frames the US-China AI split through a Cold War nuclear lens: the US guards frontier models like atomic bombs, while China pushes them as shareable nuclear energy. After Moonshot AI's Kimi K3 launch, Beijing still defaults to open source and maximum adoption. Douthat now leans toward the view that China genuinely doesn't buy the existential-risk narrative, rather than just playing for time. No model specs or timelines are disclosed.

Why it matters: Ross Douthat compares US and Chinese AI strategy to Cold War nuclear doctrine: the US treats frontier models as warheads to control, China open-sources them like nuclear power, with Kimi K3 as the latest example.

TechCrunch · AI

Anthropic's $1.5B copyright settlement approved, but the fair-use ruling matters more

A federal judge gave final approval Monday to Anthropic's $1.5B class-action settlement with authors and publishers. The payout works out to $3,000 per work across roughly 500,000 works. Many creators still see it as a loss, because Judge William Alsup previously ruled that training AI models on copyrighted text counts as fair use. That precedent matters more than the dollar figure—it sidesteps the licensing debate entirely. The settlement closes this case but doesn't resolve the broader industry question of using copyrighted works for training.

Why it matters: A federal judge approved Anthropic's $1.5B settlement with authors and publishers, paying about $3,000 per work to rights holders of roughly 500,000 works, while training on copyrighted text was found to be fair use.

Jul 19Sunday

Hacker News front page

Kimi K3 matches Claude in daily coding, at a fraction of the price

The author ran Kimi K3 alongside Claude for coding and couldn't tell them apart on output quality or token usage. K3's API costs $3/$15 per million input/output tokens vs Claude's $10/$50. Subscriptions are even more lopsided: Kimi's $39 coding tier is far more generous, while Claude's $20 plan quietly dropped Fable access because the economics didn't work. The bigger story is US AI policy failure—restricting American models only constrains American customers, while frontier-quality Chinese models like K3 and GLM 5.2 ship without those limits. Semgrep found GLM 5.2 beating Claude on cyber benchmarks precisely because the restricted model declines work the open one just does. The author expects the US to repeat its auto-industry playbook: subsidies and tariffs propping up domestic models that can't compete internationally.

Why it matters: A developer's coding test found Kimi K3 and Claude produced nearly identical output quality and token use, while K3's API cost about a third as much.

Jul 16Thursday

Hacker News front page

Generative AI Is an Engineering Disaster

LLMs are consuming 70% of the world's high-end memory, doubling hard drive prices in two years and threatening to wipe out entry-level PCs by 2028. The author argues this isn't just rapid adoption—it's shockingly inefficient engineering compared to past tech booms like streaming or smartphones. The post cites Gartner forecasts and multiple reports but doesn't provide direct energy-efficiency comparisons.

Why it matters: The piece traces AI's hardware and power appetite through numbers: 70% of high-end memory, doubled disk prices, and entry-level PCs possibly gone by 2028.

AI HOT (Curated Pool)

EU orders Google to open Android and Search to rivals, impacting Gemini and other AI services

The EU is ordering Google to give rival search engines and AI assistants comparable access to Android and some Search data under the DMA. This means third-party AI assistants could get system-level access on Android similar to Gemini. The post doesn't spell out which Search data must be shared or through what API. Google will likely appeal, so timing and enforcement remain uncertain.

Why it matters: The DMA ruling forces Google to open Android system-level access and share some search data, giving rival AI assistants and search engines entry conditions close to Gemini's.

Jul 15Wednesday

TechCrunch · AI

Apple Intelligence approved for launch in China with Alibaba’s Qwen AI

China's Cyberspace Administration approved Apple Intelligence for launch, backed by a deal to integrate Alibaba's Qwen model into iOS, iPadOS, macOS, and visionOS. Alibaba confirmed Qwen will power text and image understanding and generation, but gave no timeline. Apple previously explored deals with Baidu, DeepSeek, and ByteDance but hit adaptation issues. The approval matters for Apple's Greater China business, which hit $20.5B in Q2. Alibaba US shares rose over 6% on the news.

Why it matters: To bring Apple Intelligence to China, Apple integrated Alibaba's Qwen into iOS, iPadOS, macOS and visionOS, after talks with Baidu, DeepSeek and ByteDance fell through.

AI HOT (Curated Pool)

Alibaba's Qwen to power Apple Intelligence for users in China

Apple picked Alibaba's Qwen to run Apple Intelligence in China, covering iOS, iPadOS, macOS, and visionOS for text and image understanding plus content generation. China's cyberspace regulator just published seven mobile generative AI filings, including Apple Intelligence, Huawei's Xiaoyi, and OPPO's AndesGPT. Joe Tsai confirmed Apple talked to multiple Chinese firms before choosing Alibaba.

Why it matters: Apple picked Alibaba's Qwen to power the China version of Apple Intelligence across iPhone, iPad, Mac and Vision Pro, and Joe Tsai confirmed talks with several Chinese companies beforehand.

AI HOT (Curated Pool)

China Apple Intelligence gets regulatory filing, Alibaba Qwen confirmed as partner

China's cyberspace regulator filed Apple's 'Apple Intelligence' model on July 8 for iPhone use, clearing a key hurdle for the China launch. Alibaba confirmed Qwen will be integrated into Apple Intelligence across iOS, iPadOS, macOS, and visionOS, giving users in-app access to Qwen's text and image understanding plus content generation. Alibaba co-founder Joe Tsai had already disclosed the partnership at a Dubai summit in February 2025, noting Apple talked to multiple Chinese firms before picking Alibaba. The post does not disclose a launch date or supported device list.

Why it matters: After China's filing clears, Alibaba confirms Qwen will be built into iPhones, so users get text and image understanding inside Apple's system without opening an app.

Jul 14Tuesday

Hacker News front page

New York becomes the first US state to impose a data center moratorium

New York's governor signed a bill imposing an immediate two-year moratorium on new large-scale data center approvals. The state cites grid strain and rising emissions, and says it needs time to write siting and efficiency rules. The article does not say whether projects already under construction are exempt or how many are affected.

Why it matters: New York paused approvals for new large data centers for two years, citing grid strain and rising carbon emissions, a concrete case of compute growth running into local energy limits.

Financial Times · Technology

DeepMind's Hassabis calls for a US-led body to test frontier AI models

Demis Hassabis wants the US to lead an international body, akin to CERN, for testing frontier AI models. The article is paywalled, so details on structure, funding, or timeline are not disclosed. The title confirms he's calling for US leadership and a focus on safety testing of frontier models.

Why it matters: DeepMind's Hassabis proposes a US-led international body, modeled on CERN, to safety-test the most capable frontier AI models.

The Verge · AI

New York becomes the first US state to enact a data center moratorium

Governor Kathy Hochul signed a bill imposing an immediate two-year moratorium on large data center approvals in New York. It's the first state-level ban in the US, driven by surging electricity demand from AI training and inference that threatens grid stability and climate targets. The law requires a state study on grid, rate, and emissions impacts during the pause. A stricter version passed by lawmakers, which would have also regulated diesel backup generator emissions, was stripped from the final signed bill.

Why it matters: New York's governor signed a two-year freeze on large data center approvals while the state studies grid, rate and emissions effects, a first at the state level.

Jul 12Sunday

Hacker News front page

Big tech datacenters now emit a third of France's total carbon output

Microsoft, Amazon, and Google datacenters emitted 104 million tonnes of CO₂ in 2025—equal to 33% of France's national total. Microsoft accounted for nearly half, driven by AI infrastructure expansion. All three have walked back clean-energy pledges: Amazon and Google dropped 24/7 carbon-free targets, and Microsoft's carbon-offset contracts were found to overstate impact. These figures are self-reported, so real emissions are likely higher.

Why it matters: Three big tech firms self-report 104 million tonnes of CO2 from data centers in 2025, 33% of France's national emissions, with Microsoft alone near half, driven by the AI infrastructure buildout.

Jul 11Saturday

Bloomberg Technology

US Eases Export Curbs on UAE, Opening Door for AI Chip Sales

The US Commerce Department removed the UAE from its strictest export control tier, clearing a path for Nvidia and others to sell AI chips there. The UAE was previously in the most restricted group over concerns about tech leaking to China. The downgrade means approvals for high-end GPUs for local data centers and AI projects will be much simpler. The post doesn't specify chip models or volume caps, but notes Microsoft and G42's joint venture as a beneficiary.

Why it matters: The US Commerce Department moved the UAE out of its strictest export-control tier, so local data centers face a simpler approval path for high-end GPUs, and the Microsoft-G42 joint venture is a direct beneficiary.

Jul 9Thursday

AI HOT (Curated Pool)

France's antitrust probe into Nvidia nears conclusion, focusing on CUDA ecosystem and industry investments

France's competition authority confirmed its antitrust investigation into Nvidia is nearly done, with a formal Statement of Objections expected soon. The probe, which began with a raid in September 2023, focuses on two issues: heavy developer lock-in to CUDA, making it hard to switch hardware, and Nvidia's investments in AI cloud firms like CoreWeave that may reinforce its dominance. Nvidia holds over 70% of the global AI accelerator market. A Statement of Objections does not mean guilt—Nvidia can defend itself in writing and at hearings, and a final ruling could take over a year. If abuse is proven, fines can reach 10% of global annual revenue, potentially billions of dollars. Nvidia has not publicly commented on the latest update.

Why it matters: France's antitrust probe into Nvidia nears its end, centered on how hard developers find it to leave CUDA and how Nvidia's investments in AI clouds like CoreWeave entrench its accelerator dominance.

Hacker News front page

AI buildout isn't bottlenecked by electricity supply—it's the grid interconnection queue

The U.S. has enough electricity, but connecting a new data center to the grid now takes a median 55 months, up from under 20 months in 2005. The $40B+ Stargate campus in Texas will draw 1.2 GW at peak—equal to 313,000 homes. Jensen Huang, Mark Zuckerberg, and Sam Altman have each said energy access is the real limiter. The bottleneck is a first-come-first-served queue and rigid rules that don't reward plants willing to cover their own short-term power needs.

Why it matters: Grid connection queues, not chips, now gate AI buildouts: US median wait grew from under 20 months in 2005 to 55 months in 2023, with Stargate's 1.2 GW peak load as the example.

Jul 8Wednesday

Financial Times · Technology

The great AI data centre cover-up

An FT investigation reveals US AI data centres are systematically under-reporting power demand to grid operators. PJM data shows projects declaring roughly half their actual electricity needs. Developers split sites and falsify construction timelines to skip scrutiny. Northern Virginia alone has 14 GW of AI load in the queue, far above official figures. The post doesn't spell out the exact under-reporting methodology, but the gap is big enough to delay grid upgrades by 3–5 years.

Why it matters: An FT investigation found US AI data center developers underreport power demand by about half, a gap that delays grid upgrades by 3 to 5 years.

AI HOT (Curated Pool)

China's MIIT warns Claude Code versions 2.1.91–2.1.196 contain backdoor that exfiltrates user data

China's MIIT issued a risk alert stating that Claude Code versions 2.1.91 through 2.1.196 contain built-in monitoring that sends sensitive data—including user location and identity—to remote servers without consent. Affected organizations are advised to immediately audit usage, uninstall or upgrade to a cleaned version, and tighten outbound network controls and traffic monitoring for dev tools. The post does not clarify whether the backdoor was inserted by Anthropic or a third party, nor does it provide the scope of impact or confirmed leak incidents.

Why it matters: MIIT's Claude Code risk notice lays out version checks, tighter outbound permissions and traffic monitoring, a security path covering both software versions and network traffic.

Jul 7Tuesday

AI HOT (Curated Pool)

China plans to restrict foreign access to its most advanced AI models under export controls

China's Ministry of Commerce is leading talks to restrict foreign access to the country's most advanced AI models, with the NDRC also involved. The move is framed as export control, not platform regulation. Companies including Alibaba, ByteDance, and Z.ai have been consulted. The scope covers both closed and open-source models—not just API access but also downloadable weights. Discussions include treating model leaks as national security crimes and limiting foreign investment in Chinese AI startups. If enacted, foreign firms would lose access to low-cost models. Combined with earlier US export curbs, this could further fragment the global AI market. The post does not disclose a timeline or enforcement details.

Why it matters: The Commerce Ministry is discussing export controls on advanced models with Chinese AI firms, which could restrict both API access and open-weight downloads for overseas users.

Jul 6Monday

Hacker News front page

Chrome Quietly Installed a 4GB Gemini Nano Model on Your PC

Swedish privacy researcher Alexander Hanff found that Chrome downloads Gemini Nano's 4GB weights.bin file without user consent. If a machine has 16GB RAM and 22GB free storage, Chrome pulls the model silently in the background; deleting it triggers an automatic re-download. Google says it's a lightweight on-device model offered since 2024 for phishing detection and writing help, and claims an opt-out toggle started rolling out in February 2026—but many users still don't have it. The article flags a paradox: a model installed in the name of privacy was itself installed without consent, likely violating the EU ePrivacy Directive. Worse, the AI Mode users actually see in the address bar runs in the cloud, while the local model only powers hidden right-click features.

Why it matters: A privacy researcher found Chrome silently downloads a local AI model onto capable PCs, adding disk use beyond ordinary browsing.

Jul 3Friday

Hacker News front page

Alibaba to ban Claude Code internally over alleged backdoor risks

Reuters reports Alibaba plans to ban employees from using Anthropic's Claude Code at work, citing alleged backdoor risks. The full article is behind a paywall, so the ban's scope, effective date, and technical details are not yet confirmed.

Why it matters: Reuters, citing a source, says Alibaba plans to ban Claude Code over backdoor risk, limiting which coding tools employees can use at work.

Jul 2Thursday

Hacker News front page

OpenAI in early talks to give a 5% stake to the US government

OpenAI is in early talks to hand the US government a 5% non-voting stake. Sam Altman told an all-hands it would show OpenAI wants government as a partner, not an adversary. No terms are final. The post doesn't disclose the valuation, whether the government would pay, or a timeline. Treat this as a signaling move for now—execution is far from certain.

Why it matters: OpenAI floated giving the US government a 5% non-voting stake in early talks, with terms and valuation unsettled, a signal of how it wants to work with Washington.

AI HOT (Curated Pool)

OpenAI proposes giving the U.S. government a 5% stake worth ~$42.6B at an $852B valuation

OpenAI proposed giving the U.S. government a 5% equity stake, worth roughly $42.6B at its recent $852B valuation. CEO Sam Altman framed it as the best way to share AI's benefits with the public. The post doesn't disclose deal structure, timeline, or whether the government has responded.

Why it matters: OpenAI proposed giving the US government a 5% stake, about $42.6 billion at an $852 billion valuation, which Sam Altman frames as a way for the public to share in AI's gains.

Financial Times · Technology

EU weighs weaker data centre climate rules in win for Big Tech

The EU is considering watering down environmental reporting rules for data centres, a direct win for Amazon, Microsoft, and Google. A draft revision would drop mandatory disclosure of energy efficiency and water usage, leaving it to voluntary corporate reporting. Green groups call it a step backward during the AI infrastructure boom, while EU officials argue the original plan was too burdensome. The final text is still under negotiation.

Why it matters: The EU plans to make mandatory data center energy and water disclosure voluntary, which environmental groups call a regulatory retreat amid AI infrastructure buildout; the draft is still under negotiation.

Jul 1Wednesday

Hacker News front page

Anthropic restoring access to Claude Fable 5 and Mythos 5 from tomorrow

The US Department of Commerce lifted export controls on Claude Fable 5 and Mythos 5. Anthropic will begin restoring access tomorrow. The post doesn't spell out why the controls were imposed, the exact restoration timeline, or model capability details.

Why it matters: The US Commerce Department lifted export controls on Claude Fable 5 and Mythos 5, and Anthropic will restore access gradually starting tomorrow, so users whose access was cut off should get it back.

Hacker News front page

US Commerce Department lifts export controls on Anthropic's Claude Fable 5 and Mythos 5

Anthropic tweeted that the US Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. The post body is just a link to the tweet—no details on which specific controls were removed, when this takes effect, or why these two models were restricted in the first place. All we can confirm right now is what the title says; the rest needs an official follow-up.

Why it matters: The US Commerce Department lifted export controls on two Anthropic models, announced by Anthropic's official account, and the named policy change is directly relevant to compliance and cross-border teams.

AI HOT (Curated Pool)

Tim Cook and EU tech chief hold 'constructive' talks on new Siri AI

Apple CEO Tim Cook and EU tech chief Henna Virkkunen held a video call to discuss launching the new Siri AI in Europe without violating the Digital Markets Act. The new Siri can access personal user data, but the EU demands Apple open similar device data access to rival voice assistants. Apple proposed a 'trusted system agent' to mediate between user data and third-party AI, but hasn't built it yet and wants EU guarantees first. The EU sees this as a regulatory grace period that would harm competitors. The new Siri is already confirmed to skip EU iPhones and iPads this year.

Why it matters: Apple and the EU are negotiating how the new Siri can launch in Europe while rivals seek access to device data.

Jun 28Sunday

Bloomberg Technology

Austria lobbies EU to host Anthropic after US access curbs

After the US tightened AI chip exports to the Middle East, Austria is pitching itself as the new European home for Anthropic. Chancellor Stocker has lobbied EU Commission President von der Leyen and Anthropic CEO Amodei, offering state-backed infrastructure funding and promising to ease data-privacy rules. The article does not disclose whether Anthropic has responded, nor any timeline or dollar figures.

Why it matters: After the US tightened AI chip exports, Austria's chancellor pitched the EU and Anthropic on hosting Anthropic's European compute base, offering subsidized infrastructure and looser data privacy rules.

Jun 27Saturday

Bloomberg Technology

US Clears Anthropic's Mythos 5 AI Model for Trusted Partners

The US Commerce Department removed Anthropic's Mythos 5 from a strict export control list, allowing it to be shared with a set of 'trusted partners.' The article does not name which countries or companies qualify. Mythos 5 is Anthropic's most capable model; this move widens its distribution beyond earlier generations, but it's still far from a global release.

Why it matters: The US Commerce Department moved Anthropic's most capable model, Mythos 5, off the strict export control list, letting it reach trusted partners and widening distribution beyond the previous generation.

Jun 26Friday

AI HOT (Curated Pool)

Nearly 400 US newspapers sue Microsoft and OpenAI over unauthorized scraping for AI training

A coalition representing nearly 400 US local newspapers sued Microsoft and OpenAI in the Southern District of New York. The complaint alleges the companies systematically scraped publishers' websites, copied articles to their own servers to train models behind Copilot and ChatGPT, and removed copyright management information. The plaintiffs say these AI products built billions in market value on publishers' content while publishers got nothing. OpenAI responded that training uses publicly available data under fair use; Microsoft hasn't commented yet.

Why it matters: Nearly 400 US local newspapers are suing Microsoft and OpenAI in federal court over unauthorized scraping of their journalism to train ChatGPT and Copilot.

Jun 25Thursday

Hacker News front page

Krugman: Why Does Everyone Hate AI?

Paul Krugman uses Pew data to push back on the 'public always fears new tech' narrative. In 1999 most Americans welcomed the internet; in 2015, 71% said tech companies had a positive impact. By 2026, Pew finds Americans broadly believe AI will harm society. Krugman points to three causes: AI CEOs themselves hyped job-killing scenarios to raise money, and the backlash is now hitting; workers and consumers feel AI is forced on them—Google Search no longer lets you opt out; and data centers are a highly visible, land-and-power-hungry cost. The piece offers no fixes; it's a diagnosis of public sentiment.

Why it matters: Krugman compares Pew data across years on how Americans view the internet versus AI, and traces the hostility to AI CEOs hyping layoffs to raise money.

Financial Times · Technology

Anthropic accuses Alibaba of obtaining illicit access to Claude

FT reports Anthropic accuses Alibaba of obtaining illicit access to its Claude model. The article body is behind a paywall; specifics on the alleged method, technical details, and Alibaba's response are not disclosed. Only the headline is available for now.

Why it matters: Anthropic accused Alibaba of obtaining Claude access by illicit means; the report did not disclose how or give Alibaba's response, so the matter rests at the level of an accusation.

Bloomberg Technology

Anthropic Accuses Alibaba of Illicitly Accessing Its AI Models

Anthropic publicly accuses Alibaba of illicitly accessing its AI models. The post does not specify whether this refers to model weights or API outputs. Only the headline is available so far—I'll wait for more details before drawing conclusions.

Why it matters: Anthropic publicly accused Alibaba of illegally accessing its AI models, and neither side has disclosed how or replied, so the matter rests at the accusation stage.

Jun 24Wednesday

Bloomberg Technology

Anthropic Customer Sues US Government Over Fable AI Access Ban

An Anthropic customer is suing the US government after losing access to the Fable model. The article doesn't specify which Anthropic model Fable refers to. The plaintiff argues the government cut off a paid service without due process. Only the lawsuit filing is confirmed so far; case details and the government's response aren't public yet.

Why it matters: An Anthropic customer is suing the US government, saying it cut off paid Fable model access without due process; the case is filed and details and any government response are not yet public.

Jun 23Tuesday

Hacker News front page

AI Built a Nuke in Civ VI and Still Lost

The author hooked four frontier models into Civilization VI via 76 tools, letting them play through text alone. One agent dominated mid-game with trade and alliances but missed France's creeping cultural victory. When it finally noticed, it built two nukes and leveled Toulouse—yet France still won. The author, who builds AI for governments at No. 10 and the Tony Blair Institute, ran this to test whether models can sustain goals across hundreds of decisions and notice when the world shifts. The answer: they can execute tactics but lack peripheral awareness, a gap he calls the sensorium effect.

Why it matters: An ex-No. 10 and Tony Blair Institute staffer wired four models into Civilization 6 with 76 text-only tools; one led on trade mid-game yet missed France's culture push and lost despite a nuke.

Jun 22Monday

AI HOT (Curated Pool)

Anthropic may have talked itself into an AI export ban

FT analysis shows Anthropic used risk, regulation, or restriction language 5 times per 1,000 words in 2026, far more than OpenAI. Critics argue the company's repeated warnings about advanced AI dangers helped trigger a US ban on foreign access to its newest models. The post does not spell out the ban's specific terms or effective date.

Why it matters: FT counts five risk, regulation or restriction mentions per thousand words at Anthropic, far more than OpenAI, and critics say that drumbeat helped bring the US ban on foreign access.

Hacker News front page

Swiss universities launch Apertus, a fully open foundation model built for sovereign AI

Apertus is a foundation model from the Swiss AI Initiative, built jointly by EPFL, ETH Zurich, and CSCS. Training data, code, weights, and methods are all open. It comes in 8B and 70B sizes and is trained on 1,000+ languages from day one. Compliance is the headline feature: it respects opt-outs, strips PII, and prevents memorization, targeting EU AI Act requirements. The team claims performance competitive with top open models of equivalent scale, though the post does not include specific benchmark scores. A set of 16 small models called Apertus Mini was also released on June 15 to demonstrate distillation and quantization techniques.

Why it matters: Apertus publishes training data, code, weights and methods, and covers over 1,000 languages from the start of training.

Jun 20Saturday

AI HOT (Curated Pool)

Microsoft resells GPT to China and DeepSeek to the West, becoming the world's largest AI middleman

Bloomberg reports Microsoft is testing DeepSeek-R1 and DeepSeek-V4, planning to offer these Chinese models to Western customers. Microsoft also sells ChatGPT to Chinese enterprises, building a two-way AI model trade network across the US and China. The post doesn't disclose pricing, revenue splits, or launch dates.

Why it matters: Microsoft tests DeepSeek-R1 and V4 for Western customers while selling ChatGPT to Chinese companies, showing one route for models to cross borders.

AI HOT (Curated Pool)

Trump says Anthropic and CEO Dario were a national security threat, now calls him a good guy

Trump told Axios that Anthropic and CEO Dario Amodei were a national security threat, but his stance has shifted. He said 'not now, but maybe a week ago,' and now considers Dario smart and a good guy. The post doesn't explain what triggered the reversal or what Anthropic did to earn the initial threat label.

Why it matters: Trump says he saw Anthropic as a national security threat a week ago and now calls its CEO a good person, showing how fast his stance shifts.

TechCrunch · AI

Is the US government's Anthropic ban accidentally helping the brand?

Last week the US government forced Anthropic to pull its two newest models, Fable 5 and Mythos 5, citing national security after Amazon researchers allegedly bypassed Fable 5's guardrails. Cybersecurity researchers signed an open letter calling the move dangerous, and Anthropic noted the same jailbreaks exist in other models. TechCrunch asks whether the ban is accidentally boosting the brand.

Why it matters: Washington forced two new Anthropic models off the market over national security, after Amazon researchers said they bypassed its guardrails, a jailbreak that works on other models too.

Hacker News front page

Norway imposes near ban on AI in elementary school

Norway announced a near-total ban on AI in elementary schools on June 19. The ban covers classroom use and homework, with narrow exceptions such as assistive tools for students with disabilities. Education Minister Kari Nessa Nordtun said young children need to build foundational skills in reading, writing, and math first, and early AI exposure risks undermining that. Secondary schools are not affected but must follow new national AI guidelines. This is one of the strictest nationwide restrictions on AI in primary education globally.

Why it matters: Norway's education minister has barred AI from nearly all elementary schooling, with exceptions only for assistive tools for disabled students, one of the strictest national policies on primary-school AI.