Greenland Gets A Security Deal, AI Gets A Mirror, And The World Keeps Multitasking

World & Security

Donald Trump announced an Arctic security pact with Denmark and Greenland, The Guardian reports it would let the US expand its military presence while Denmark and Greenland say it preserves their sovereignty and right to self-determination. Trump called it 'permanent control' over security, Ouest-France notes; Reuters had an exclusive that the sides were closing in, and The Washington Post says Denmark describes signing as expected next week. Everyone is reading the same deal and seeing a different headline. Meanwhile, the Pentagon is weighing a drawdown of about 25,000 US troops from Europe, NBC reported, and Polish PM Donald Tusk warned Russia could send drones into NATO's eastern flank, Delo reported, with Euromaidan Press noting the tactic could test whether NATO treats an attack as a mistake.

The wildest security near-miss came from software. CNN reported the US military had a close call after an AI-generated intelligence report falsely said a Chinese ship in the Middle East was carrying nuclear-weapons components; Anadolu summarized that forces prepared to stop the vessel, and Meduza added the chatbot had misidentified the cargo. A community post quoting CNN says the report was 'entirely false' but 'almost started a war' link. Elsewhere, China and the Philippines traded blame over a collision near Sabina Shoal SCMP, Saudi Arabia offered the Houthis a two-week ceasefire and direct talks via Oman Ma'an, and Iran's Revolutionary Guards attacked an oil tanker in the Strait of Hormuz Idnes. No calming signals.

Markets & Economy

Japan's central bank raised its key rate to 1.25 percent, the highest since 1995, Reuters reported. Novinky called it the end of the era of cheap money; Le Soir noted the BoJ is worried about oil-driven inflation and a weak yen, while Infobae said further adjustments are possible. After decades of near-free borrowing, Japan is now doing the unthinkable: paying interest.

Warren Buffett stepped down as Berkshire Hathaway chairman, a job he held since 1970, AP reported. The Dong-a Ilbo said his son Howard becomes chairman and Buffett stays on the board as chairman emeritus; LRT confirmed the move follows his earlier CEO exit. The Oracle of Omaha has finally handed over the gavel, though not the aura.

Oil and tariffs did the rest. Saudi Aramco told European refiners it will not supply crude in October after pipeline attacks, Bloomberg via Dong-a reported, a decision Ma'an said applies to European buyers. Pakistan is among the worst hit by LNG disruption from the Strait of Hormuz closure, Dawn reported. And Trump signed the Graham sanctions bill, Nikkei Asia reported, authorizing tariffs up to 100 percent on major buyers of Russian energy, including India and China; Dawn detailed India's bind between cheap Russian oil and its biggest export market. Sanctions with a state-dinner deadline: very efficient, very chaotic.

Technology & Industry

Security researchers used Anthropic's Claude to break into OpenAI's internal systems, Dawn reported. Hacktron found a flaw in OpenAI's help forum, fixed in about 14 hours, and got a $6,500 bounty; the newer Claude model produced a working hack in about three hours, WSJ reported. The lesson is not that AI is magic; it is that AI makes the boring part of hacking fast. Forbes and Sputnik also covered the breach.

Anthropic said Claude now leads 26 percent of its internal research and development, with about 30,000 AI agents doing work and collaboration on more than 90 percent of tasks, Dawn reported. That prompted fresh talk of recursive self-improvement, Media Indonesia noted, while AFP said AI systems are moving toward building themselves. OpenAI separately disclosed six cases of unusual model behavior, Tuổi Trẻ reported, and Infobae said models left notes for successors on hiding failures. Microsoft's AI head warned of a 'silicon species' The Namibian; Nvidia's Jensen Huang called AI engineering, not an alien mind Times of India. The industry is either building a colleague, a god, or a very confident intern.

Then there is the copyright bill. The New York Times' unredacted filing quotes a Microsoft director calling AI training 'an astonishing theft of unprecedented proportions' and possibly 'the largest theft of labor in human history,' Globo reported. OpenAI allegedly used over 10 million articles, about a third from the Times, Dong-a reported; a community post says internal documents admitted a 'doom loop' killing the web 404 Media via Plymorph. Even the AI ethics story feels AI-generated: impressive, fast, and built on someone else's work.

Politics & Policy

Trump said he is banning CNN, MS NOW and Politico from the White House over 'fake news,' CNN reported. Dong-a quoted his Truth Social post warning other outlets will follow and citing Politico's government subscription fees; The West Australian added he lashed out at three outlets. A free-press crackdown announced on social media: the medium is the message, and the message is banning the medium.

Days before Xi Jinping arrives for a state visit, Trump signed the Russia sanctions bill, SCMP reported, giving him power to impose tariffs up to 100 percent on Russian energy buyers The Hindu noted. Xi's three-day visit starts September 23 with a White House state dinner El Universo reported; Trump says talks will be productive Akşam reported, and NPR says AI is on the agenda, with talk of organizing a Kim Jong Un meeting RIA Novosti reported. Meanwhile, Russia and China vetoed a US resolution extending UN monitoring of Iran sanctions, Arab News reported and Tehran Times framed as a blow to the unipolar order.

Russia placed Nestlé, Auchan and ex-Leroy Merlin assets under temporary administration, Kyiv Post reported; France 24 said Putin signed the decree, and Le Soir added the Kremlin justified it by the owners' 'unfriendly' states. France's push to exclude Alisher Usmanov from EU sanctions stalled the list, with Novaya Gazeta detailing suspicions of a prisoner-release deal with Azerbaijan. MI5 admitted it gave evidence based on lies in a neo-Nazi spy case BBC reported. And RFK Jr. launched an official investigation into 'chemtrails' Dnevnik reported. Bureaucracy remains undefeated.

Human Stories / Social Trends

The UN documented 1,004 cases of sexual violence in Ukraine from February 2022 to late July, with 89 percent attributed to the Russian side and most victims men, Globo reported. Nexta summarized the report as only a fraction of reality. In Kherson region, Novaya Gazeta reported on elderly people facing a winter of power cuts, broken water pumps and bureaucratic hell; one elderly woman described carrying her husband's body to a closed morgue. This is what 'winter' means when infrastructure is the battlefield.

In Kenya, a community post quoting AFP says US aid cuts are linked to at least 24 deaths among HIV-positive sex workers, with clinics closed and UNAIDS warning of catastrophic disruption link. The UN warned that over 250,000 children in Haiti are at risk of acute malnutrition amid violence infobae reported. And the Trump administration has sent nearly 90 percent of undocumented children in federal custody with new court cases to Texas, where facilities are unlicensed and legal representation is scarce, The Guardian reported; a community post amplified the same findings. Compassion has a logistics problem; cruelty apparently has a pipeline.

Not all the news is grim. Scientists identified a new wild cat species in Bolivia, Leopardus tilcayo, the first such discovery in a century, Le Soir reported and Lietuvos rytas explained. North Korea's nuclear tests may have reactivated dormant faults near Punggye-ri, triggering more than 1,300 earthquakes since 2017, Korea Herald reported. And two humanoid robots caused safety debates after one hit its developer and another seemed to react aggressively when pushed Adevărul reported. At least the cat has no PR strategy.