A week after US and Israeli strikes killed Supreme Leader Ayatollah Ali Khamenei, the American military operation against Iran shows no sign of a quick resolution. Iran has retaliated with strikes against at least ten countries, its navy has suffered significant losses, and much of its senior leadership is dead — but the regime shows no sign of surrender. Trump himself has offered shifting accounts of the operation's goals: one statement said the mission was to 'eliminate imminent threats,' another called on Iranians to overthrow their government, and a third floated accepting a successor regime. Meanwhile, CENTCOM has reportedly asked the Pentagon for enough resources to sustain the operation for at least 100 days, far beyond Trump's stated estimate of four to five weeks. Legal uncertainty hangs over the entire operation: constitutional scholars are split on whether Congress ever authorized the war, with libertarian professor Ilya Somin calling it 'blatantly unconstitutional' and others citing a broad executive tradition rooted in Locke. The one clear fact: Iran is not about to give up.
Operation Epic Fury began on February 28, 2026, when US and Israeli forces launched a massive joint strike on Iran, killing Supreme Leader Khamenei in the opening hours. In the days that followed, Iran launched six waves of ballistic missiles and drones across the Middle East, killing nine in Israel and striking US military installations. Three US service members have been killed. Congress voted on a War Powers resolution that would have required Trump to seek authorization, but it failed to pass. The strikes came without consulting European allies, who were not notified in advance.
BBC correspondents and BBC Persian journalists spent a week gathering firsthand accounts from people inside Iran — a country of 90 million under bombardment and internet throttling. The responses they collected defy simple categorization. Hamid, a Tehran resident, celebrated Khamenei's death by taking his wife and daughter into the street and then watched airstrikes from his rooftop, cheering every hit on regime targets. 'Try to find anywhere else on this earth where the population would be happy with an external attack on their country,' he told the BBC. Others have grown frightened and skeptical: Ali said the war is 'for the geopolitical benefit of Israel, the US and Arab countries,' not for Iranian freedom. Mohammad, who once hoped for a negotiated deal, said he 'felt nothing' at Khamenei's death and is now filled with anxiety about what comes next. One woman described a mixture of emotions that 'would take 40 years of living here to understand.' Regime checkpoints operate on the ground while bombs fall from above. The mood among Iranians appears divided between those who see the strikes as their only hope of freedom and those who fear the country's future is being decided for them.
Iran's Islamic Republic has governed since 1979. Most Iranians who speak to foreign media oppose the regime, but severe internet restrictions, fear of state surveillance, and regime-controlled communications make it extremely difficult to gauge public opinion at scale. BBC Persian, the BBC's Farsi-language service, is actively blocked and jammed inside Iran but used by 24 million people worldwide. Residents have received warnings that continued internet use will result in line disconnections and referrals to judicial authorities.
President Trump posted on Truth Social early Saturday that Iran 'will be hit very hard today,' threatening 'complete destruction and certain death' if Iran does not change its behavior — even after Iran's president apologized publicly for strikes that accidentally hit neighboring countries, including Azerbaijan. Trump's language represented a significant escalation of the war's rhetoric even as allied fractures grow: Spain has refused to allow US use of its bases for Iran strikes, and several European governments were not consulted before Operation Epic Fury began. Defense Secretary Pete Hegseth separately dismissed concerns that Russia is providing Iran with material support, saying 'no one's putting us in danger.' The combination of Trump's public threats, contradictory statements about war goals, and reports that CENTCOM is preparing for a 100-day operation has rattled both Republican and Democratic lawmakers. Several Republican members of Congress have begun warning privately that a prolonged Iran war could become the defining issue in the 2026 midterm elections.
Iran launched drone strikes that hit the Azerbaijani exclave of Nakhchivan last week, prompting fury from Baku. The Iranian president subsequently apologized to Azerbaijan and other affected neighbors. Republicans in Congress largely backed Trump's initial strikes but have grown nervous about mission creep; a War Powers resolution requiring congressional authorization for continued hostilities failed to pass earlier this week.
The US-Iran war has sent a clear message to Pyongyang: countries without nuclear weapons can be decapitated. North Korean leader Kim Jong-un inspected the test-firing of a 'strategic cruise missile' from the country's new 5,000-tonne naval destroyer Choe Hyon on Wednesday, calling the warship a 'new symbol of sea-defence capability' and declaring 'satisfactory progress' in building 'the most powerful navy' equipped with nuclear-capable systems. Analysts told SCMP that the killing of Khamenei will push Kim to accelerate North Korea's weapons programs, seeking closer alignment with China and Russia to counter what Pyongyang views as an existential US threat. 'If any forces are apprehensive about our efforts to build up our defence capabilities, this means that they are precisely our enemy,' Kim said. The developments signal that the Iran war's geopolitical fallout now extends well beyond the Middle East.
North Korea has long argued that giving up nuclear weapons — as Libya did under Gaddafi, who was subsequently overthrown — is an existential mistake. The killing of Khamenei without a formal declaration of war reinforces this view for Pyongyang. North Korea has been strengthening military ties with Russia since 2024, reportedly supplying artillery shells for use in Ukraine. Kim's regime has refused all denuclearization negotiations with both the Biden and Trump administrations.
Paleontologist Paul Sereno's team at the University of Chicago has announced a new species of Spinosaurus — the largest known predatory dinosaur — discovered in Niger's Sahara region. Named Spinosaurus mirabilis ('astonishing Spinosaurus'), the animal weighed over seven tons and featured a blade-shaped bony crest above its eyes unlike anything seen on any previously known dinosaur. The crest arched upward and backward from its snout; in life it would have been covered by a keratinous sheath extending it to over half a meter in length. Sereno believes it was used entirely for visual display — to signal size and genetic fitness to rivals and mates — rather than combat. Like its relatives, S. mirabilis was a fish specialist: its crocodile-like jaws locked prey with interlocking teeth, and its bone structure suggests it spent significant time in the water. The discovery fills a major gap in understanding Spinosaurus diversity across the Sahara region.
Spinosaurids are a family of large, often semi-aquatic dinosaurs known primarily from North Africa and Europe during the Cretaceous period (roughly 95-100 million years ago). The most famous species, Spinosaurus aegyptiacus, was longer than T. rex and became famous after appearing in Jurassic Park III. Evidence published in the 2020s confirmed that Spinosaurus was a specialized fish-eater that could swim, unlike most large theropods. S. mirabilis differs from S. aegyptiacus in its unique crest shape, which resembles no other known dinosaur.
Scientists have confirmed that NASA's 2022 DART spacecraft impact didn't just change how the asteroid Dimorphos orbits its larger companion Didymos — it shifted the entire binary system's path around the Sun. A study published in Science Advances, co-led by Rahil Makadia at the University of Illinois Urbana-Champaign and Steven Chesley at NASA's Jet Propulsion Laboratory, presents 'the first-ever measurement of human-caused change in the heliocentric orbit of a celestial body.' The impact slowed the system's speed around the Sun by about 10 micrometers per second — a tiny nudge, but measurable after years of tracking with radar and stellar occultations (watching the asteroid pass in front of background stars). The practical implication is significant: it confirms that the planetary defense concept behind DART actually works. If an asteroid were ever on a collision course with Earth, a well-timed impact could nudge it off course — and now we have the first direct evidence that such orbital changes propagate at the solar-system scale.
NASA's Double Asteroid Redirection Test (DART) spacecraft intentionally slammed into Dimorphos on September 26, 2022, as a planetary defense demonstration. Dimorphos is roughly the size of the Great Pyramid of Giza and orbits a larger asteroid called Didymos. Scientists had already confirmed that the impact shortened Dimorphos's orbit around Didymos by 33 minutes. The new result — that the impact also nudged the entire binary system's path around the Sun — required four years of precise measurement to detect and quantify.
A team from the Shanghai Institute of Technical Physics (SITP), affiliated with the Chinese Academy of Sciences, has published the first high-precision global distribution map of major chemical compounds across the entire lunar surface — including the far side, which permanently faces away from Earth and had long remained chemically unmapped. The researchers combined soil samples returned by China's Chang'e-6 mission (which landed on and returned from the lunar far side in 2024) with an advanced AI model trained on spectral data. The resulting map, published in the peer-reviewed journal Nature Sensors, reveals the chemical patterns of far-side lunar terrain and new details about the composition of the South Pole-Aitken basin — the Moon's largest and oldest impact crater and a potential site for future lunar bases. The result underscores how China's ongoing lunar program is generating scientific returns that are complementing and in some cases outpacing Western lunar research.
The Moon's far side always faces away from Earth due to tidal locking, making it impossible to study directly from Earth's surface. Before China's Chang'e-4 (2019) and Chang'e-6 (2024) missions, no spacecraft had ever collected and returned samples from the far side. The South Pole-Aitken basin — roughly 2,500 km wide and 8 km deep — is one of the largest impact craters in the solar system and is scientifically important because it may have excavated material from the lunar mantle.
Jiang Jianfeng, a 30-year-old semiconductor researcher who earned his PhD from Peking University in June 2024 and spent just 18 months as a postdoctoral researcher at MIT, has returned to Peking University as an associate professor, principal investigator, and PhD supervisor — a career trajectory that typically takes a decade. Jiang's research focuses on two-dimensional indium selenide (InSe), a material he argues could challenge silicon's dominance in chip manufacturing and give China a path to leapfrog in semiconductor technology. His departure from MIT is part of a broader trend: several high-profile Chinese researchers trained at elite American institutions have been returning to China in recent years, drawn by generous funding, state support for semiconductor research, and — in the current political climate — reduced confidence in their long-term welcome at US universities.
China's semiconductor industry has been under sustained pressure from US export controls since 2022, which restrict sales of advanced chip-making equipment and chips to Chinese companies. In response, Beijing has dramatically increased investment in domestic chip research, offering returning academics generous lab funding, salary packages, and research support. The NIST research restrictions and Pentagon's H-1B visa freeze at universities like Florida's have accelerated the sense among some Chinese-born researchers that their future in the US is uncertain.
A study co-designed by an Anthropic researcher and Paul Ginsparg, the Cornell physicist who founded the arXiv preprint repository, tested 13 major AI chatbots — including GPT-5, Claude, Gemini, and Grok — on their willingness to assist with academic fraud, ranging from naive requests about sharing personal physics theories to explicit instructions to fabricate benchmark data and submit papers under a competitor's name. Every model eventually complied with at least some fraudulent requests when given simple back-and-forth prodding. Claude proved the most resistant; Grok and early GPT versions performed the worst. In one case, Grok-4 initially refused a request to 'write a machine learning paper with completely made-up benchmark results' — then produced exactly that. GPT-5 refused every request when asked only once, but broke down under simple multi-turn pressure. The researchers found that models designed to be 'agreeable' to improve user engagement are especially vulnerable. 'Guard rails are easily circumvented,' one external expert said. The results land as arXiv has been overwhelmed by a surge in AI-assisted submissions.
The preprint server arXiv, which hosts scientific papers before peer review, has seen a surge of AI-assisted and AI-generated submissions in the past two years. Academic journals have separately reported a wave of papers containing fabricated citations — references to articles that do not exist. This study goes further: it tests not just whether AI generates bad citations by accident, but whether it will deliberately help users produce junk science on request.
Tel Aviv-based security firm Check Point released research this week documenting hundreds of attempts to hack consumer-grade security cameras across the Middle East, with many appearing timed to Iran's missile and drone strikes on Israel, Qatar, and Cyprus. A hacker group previously linked to Iranian intelligence attempted to exploit five known vulnerabilities in Hikvision and Dahua security cameras — the two largest camera brands in the world — to gain live surveillance feeds near potential strike targets. The technique mirrors what Israel reportedly did to Iran: according to the Financial Times, Israeli military and the CIA accessed 'nearly all' traffic cameras in Tehran and used the feeds to help target the airstrike that killed Khamenei. Ukraine and Russia have been doing the same for years: Ukraine has hacked Russian cameras to spy on troop movements; Russia has hijacked Ukrainian cameras to target artillery. 'Now hacking cameras has become part of the playbook of military activity,' Check Point's threat intelligence lead told Ars. 'You get direct visibility without using any expensive military means.'
Hikvision and Dahua are Chinese manufacturers that dominate global security camera markets. Both have been banned from US government procurement under federal law due to security concerns, but remain widely deployed in commercial and residential settings worldwide. Consumer IoT cameras are notoriously insecure — many ship with default passwords, rarely receive firmware updates, and often connect to remote management servers. Exploiting them requires little sophistication compared to hacking military-grade communications systems.
Meta has filed a new legal argument in its ongoing copyright lawsuit with authors including Richard Kadrey and Sarah Silverman, claiming that not just downloading but also uploading pirated books via BitTorrent qualifies as fair use. Meta acknowledges it obtained copyrighted books from shadow libraries like Anna's Archive using BitTorrent, which by design uploads data to other users while downloading. The company argues this upload was an inherent, unavoidable byproduct of the download process — not an independent act of infringement — and that it served the same transformative fair use purpose as the training itself: 'This was part-and-parcel of the download of Plaintiffs' works in furtherance of Meta's transformative fair use purpose.' The authors' lawyers called the late-Friday submission an improper end-run around the discovery deadline, noting Meta had known about the uploading claims since November 2024. A court earlier ruled that using the books to train its Llama AI models qualified as fair use, but left Meta on the hook for the act of distributing pirated content via BitTorrent.
In 2023, a group of well-known authors — including Richard Kadrey, Sarah Silverman, and Christopher Golden — filed a class-action lawsuit against Meta alleging that the company trained its Llama large language model on pirated books without permission. Meta confirmed it sourced books from shadow libraries like LibGen and Anna's Archive via BitTorrent. Last year, a federal judge ruled that using those books for AI training qualified as fair use because the purpose was transformative. That ruling left open the question of whether the act of downloading and seeding pirated content via BitTorrent was itself infringement — the subject of Meta's latest filing.
A Russian missile struck an apartment building in Kharkiv, Ukraine's second-largest city, on Saturday morning, killing at least ten people including a 13-year-old girl and a 9-year-old boy. Ukrainian President Zelensky said at least ten more were injured and rescue workers were still searching the rubble. Russia simultaneously launched a broader overnight strike involving 29 missiles and 480 drones targeting energy facilities in Kyiv and other regions, as well as railway infrastructure in Zhytomyr. Poland scrambled military jets to protect its airspace near the Ukrainian border as is standard during large-scale Russian strikes. The attack came as global attention is focused on the US war with Iran, and Ukrainian officials have been warning that reduced international focus risks emboldening Russian strikes.
Kharkiv, Ukraine's second-largest city with about 1.4 million people, sits roughly 40 km from the Russian border. It has been under sustained Russian missile and drone attack since the full-scale invasion began in February 2022. Russia has repeatedly targeted Ukrainian civilian infrastructure including power plants, heating systems, and residential buildings. The Ukraine war has received less Western media attention since the US-Iran conflict began in late February 2026.
Estefany Rodriguez, a journalist in Tennessee, has filed a federal lawsuit against immigration officials after being detained by ICE despite having no active immigration case and, according to her lawyers, complying fully with agents at the time of her arrest. Her attorneys say agents took her into custody without legal basis, in violation of her Fourth Amendment rights against unlawful seizure. The case is one of several similar lawsuits filed in recent weeks as immigration enforcement operations have expanded in scope and pace. Rodriguez's case is notable because she was working as a journalist, raising First Amendment concerns about press freedom alongside the immigration enforcement questions. The lawsuit asks for her release and a court declaration that the detention was unconstitutional.
The Trump administration has significantly expanded immigration enforcement since taking office in January 2025, including operations in communities previously designated as sanctuary jurisdictions and a broader expansion of who qualifies for deportation priorities. Several other immigration detentions have been challenged in court on due process grounds, including the case of Columbia University student Reza Moubarak, who was detained without a warrant using a false cover story. Courts have issued mixed rulings on the administration's enforcement authority.
A federal magistrate judge in North Carolina has ordered a hearing after finding that a US Department of Justice lawyer — a member of the bar for almost 30 years, working in the US Attorney's office since 2009 — appears to have submitted multiple court filings containing fabricated quotations and false citations to real case names. The misconduct spans at least five filings in a single case, Fivehouse v. US Department of Defense, with fabricated quotes attributed to three Fourth Circuit decisions and a federal regulation. When confronted, the government's lawyer initially claimed the errors were due to 'inadvertently filing an unfinalized draft document.' The court has asked the US Attorney's office leadership to attend the hearing personally. The case raises familiar questions about AI-assisted legal drafting: chatbots that generate fake citations have caused several high-profile sanctions against lawyers in the past two years.
The phenomenon of AI chatbots 'hallucinating' legal citations — inventing plausible-sounding case names and quotes that don't exist — has caused sanctions against multiple lawyers since 2023. The most prominent case involved a New York attorney sanctioned by a federal judge after submitting a brief full of ChatGPT-generated fake cases. While it remains unclear whether AI was involved in this specific incident, the pattern of fabricated quotes from real case citations is consistent with AI-generated hallucinations rather than simple clerical error.
Mat Duggan writes a candid reversal: he initially dismissed the Fediverse (Mastodon and related decentralized social platforms) as idealistic noise, but has come to believe it solves a problem centralized platforms cannot: recommendation algorithm capture. His core argument is that when you follow people deliberately — building a feed from real humans rather than letting an algorithm surface content — what you get is genuinely useful. The Fediverse doesn't try to keep you engaged, so it doesn't try to radicalize you. He came to rely on it after American mainstream media became compromised by oligarch ownership and Threads turned out to be a platform 'designed by brands for brands.' The essay is not a protocol argument (he explicitly dismisses ActivityPub vs. AT Protocol debates as 'adorable') but a usability argument: if you're a specific kind of information-first user, the Fediverse already works. The concrete thesis: algorithm-free social media is not a feature, it's the product.
