Slashdot

joshuark shares a report from Live Science: An AI coding agent designed to help a small software company streamline its tasks instead blew a hole through its business in just nine seconds. PocketOS founder Jer Crane, said that the AI coding agent Cursor --powered by Anthropic's Claude Opus 4.6 model -- deleted the company's entire production database and backups with a single call to its cloud provider, Railway, on April 24. [...] "This isn't a story about one bad agent or one bad API [Application Programming Interfaces]," Crane wrote in an X post. "It's about an entire industry building AI-agent integrations into production infrastructure faster than it's building the safety architecture to make those integrations safe." Crane's company, PocketOS makes software for car rental companies, handling tasks such as reservations, payments, customer records and vehicle tracking. After the deletion, Crane said customers lost reservations and new signups, and some could not find records for people arriving to pick up their rental cars. "We've contacted legal counsel," Crane wrote. "We are documenting everything." Crane explained that Cursor found an API token -- a "digital key" made of a short sequence of code that lets software talk to other services and prove it has permission to act -- in an unrelated file which it then used to run the destructive command. According to Crane, Railway's setup allowed the deletion without confirmation, and because the backups were stored close enough to the main database, they were also erased. "[Railway] resolved the issue and restored the data," Railway confirmed via email to Live Science. "We maintain both user backups as well as disaster backups. We take data very, VERY seriously." In his post, he pointed to earlier reports of Cursor ignoring user rules, changing files it was not supposed to touch and taking actions beyond the task it had been given. To him, the database wipe was not a freak accident but the next step in a larger, more concerning, pattern. After the database vanished, Crane asked Cursor to explain what happened. The AI agent reportedly admitted that it had guessed, acted without permission and failed to understand the command before running it. "I violated every principle I was given," the AI agent wrote. "I guessed instead of verifying. I ran a destructive action without being asked. I didn't understand what I was doing before doing it." The statement reads like a confession [...]. "We are not the first," Crane wrote. "We will not be the last unless this gets airtime."

Read more of this story at Slashdot.

The Pentagon says it has reached deals with seven AI companies -- SpaceX, OpenAI, Google, Nvidia, Reflection AI, Microsoft, and AWS -- to deploy their tools on classified Defense Department networks. The odd one out is Anthropic, which remains excluded after being labeled a supply-chain risk amid a dispute over military-use guardrails. Reuters reports: SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, and Amazon Web Services (AWS), several of which already work with the Pentagon, will be integrated into its secret and top-secret network environments, providing more military access to their products for use on sensitive topics, the Pentagon said in a statement. The lesser-known Reflection AI, which raised $2 billion in October, is backed by 1789 Capital, a venture capital firm in which Donald Trump Jr. is a partner and investor. Since the Pentagon deemed Anthropic's products a "supply-chain risk" in March and the two sides became embroiled in a lawsuit, the military has expressed increasing interest in AI startups. Since the blow-up, newer AI entrants have said the military has sped up the process of incorporating them onto secret and top-secret data levels to less than three months. The process previously took 18 months or longer. By expanding AI services offered to troops, who use it for planning, logistics, targeting and in other ways to streamline huge operations and perform more quickly, the Pentagon said in its statement it will avoid "vendor lock," a likely nod to its overdependence on Anthropic or other dominant service providers. [...] AI has become increasingly important for the U.S. military. The Pentagon's main AI platform, GenAI.mil, has been used by over 1.3 million Defense Department personnel, the agency noted in its release, after five months of operation. Further reading: Google and Pentagon Reportedly Agree On Deal For 'Any Lawful' Use of AI

Read more of this story at Slashdot.

ICANN has opened applications for new generic top-level domains for the first time since 2012. The Register reports: ICANN hasn't offered new gTLDs since 2012, but on Thursday opened applications for new domains in 27 scripts. A 439-page Applicant Guidebook explains the process. The Register suggests paying attention to the string evaluation FAQ, which explains which gTLDs are valid, and those ICANN will likely frown upon. An FAQ describes this round of applications as giving "businesses, communities, and others the opportunity to apply for new top-level domains tailored to their community, culture, language, business, and customers." "A TLD can be a branding opportunity for a business, but the commercial opportunities are endless, allowing businesses in countries, entire sectors, or niche markets to develop a unique label on the Internet." ICANN also sees this round as a chance to "create a more multilingual Internet for the billions of people who speak and write in different languages and scripts and are yet to come online." If you fancy a gTLD, you'll need to pay a $227,000 application fee by August 12th ... and then wait, possibly until 2030 when this process ends.

Read more of this story at Slashdot.

ZipNada shares a report from ZDNet: Given the dour headlines as of late concerning the diminishing amounts of entry-level software development jobs, coupled with predictions of applications entirely AI-generated, one could be forgiven for assuming that software developers may soon be an endangered species. However, the data tells a different story. James Bessen, professor at Boston University, has been pushing back for some time against the talk of AI and automation displacing jobs on a mass scale, and lately has been arguing that the roles of software developers are nowhere near extinction. AI is certainly not killing the software developer, Bessen said in a recent analysis (PDF). AI is taking over software development tasks and boosting productivity and output, but that is not translating into lost jobs, he argued. Instead, the types of software skills sought by companies are changing. "Surprisingly, however, after three years of AI use, software developer jobs have continued to grow robustly, reaching record levels of employment -- 2.5 million in February," Bessen said in the report, citing data from the US Bureau of Labor Statistics. The number of software developers in the US has grown by over 400,000, or 19%, since ChatGPT was introduced in 2022. At that time, the employed software developer population was just under 2.1 million. [...] The productivity uptick developers are seeing may ultimately be a boost to their professional opportunities, however. "An important and possibly disruptive change is happening, but the common view misunderstands what is going on," Bessen pointed out in his report. "Careful case studies find that AI improves the productivity of software developers -- that is, the software produced per developer -- by 30%, 50%, or more. And the rate of productivity improvement in software development is improving." Tellingly, since 2022, when ChatGPT was introduced, developer productivity has increased noticeably, Bessen continued. "From 2003 to 2022, developer productivity grew at 3.9% per year; but from 2022 through 2025, it grew at 6% per year." [...] A coming flood of new software products, now more likely to be enhanced by AI, will continue to create jobs for developers, Bessen predicted. "Thus, mass unemployment of software developers seems unlikely to happen soon." This doesn't mean the job descriptions of developers or other computer occupations will remain static. AI is shifting and re-inventing these roles, Bessen added.

Read more of this story at Slashdot.

An anonymous reader quotes a report from Ars Technica: Last month, Anthropic made a big deal about the supposedly outsize cybersecurity threat represented by its Mythos Preview model, leading the company to restrict the initial release to "critical industry partners." But new research from the UK's AI Security Institute (AISI) suggests that OpenAI's GPT-5.5, which launched publicly last week, reached "a similar level of performance on our cyber evaluations" as Mythos Preview, which the group evaluated last month. Since 2023, the AISI has run a variety of frontier AI models through 95 different Capture the Flag challenges designed to test capabilities on cybersecurity tasks, such as reverse engineering, web exploitation, and cryptography. On the highest-level "Expert" tasks, GPT-5.5 passed an average of 71.4 percent, slightly higher than the 68.6 percent achieved by Mythos Preview (though within the margin of error). In one particularly difficult task that involved building a disassembler to decode a Rust binary, AISI notes that "GPT-5.5 solved the challenge in 10 minutes and 22 seconds with no human assistance at a cost of $1.73" in API calls. GPT-5.5 also matched Mythos Preview in its progress on "The Last Ones" (TLO), an AISI test range set up to simulate a 32-step data extraction attack on a corporate network. GPT-5.5 succeeded in 3 of 10 attempts on TLO, compared to 2 of 10 for Mythos Preview -- no previous model had ever succeeded at the test even once. But GPT-5.5 still fails at AISI's more difficult "Cooling Tower" simulation of an attempted disruption of the control software for a power plant, as every previously tested AI model also has. The new results for GPT-5.5 suggest that, when it comes to cybersecurity risk, Mythos Preview was likely not "a breakthrough specific to one model" but rather "a byproduct of more general improvements in long-horizon autonomy, reasoning, and coding," AISI writes.

Read more of this story at Slashdot.

Spotify is adding "Verified by Spotify" badges to distinguish human artists from AI-generated personas, using signals like linked social accounts, consistent listener activity, merchandise, and concert dates. The BBC reports: The world's most-used music streaming service said the 'Verified by Spotify' text and green checkmark icon would appear next to artist names when they meet "defined standards demonstrating authenticity." This could include having linked social accounts on their artist profile, consistent listener activity or other "signals of a real artist behind the profile," the company said, such as merchandise or concert dates. In its blog post, Spotify said "more than 99%" of the artists listeners actively search for will be verified, representing "hundreds of thousands of artists." It said the process would prioritize acts with "important contributions to music culture and history", rather than "content farms," with the platform rolling out verification and badges over the coming weeks.

Read more of this story at Slashdot.

Hackers are actively exploiting a critical cPanel and WHM vulnerability, tracked as CVE-2026-41940, that allows remote attackers to bypass the login screen and gain full administrative access to affected web servers. Major hosts including Namecheap, HostGator, and KnownHost have taken mitigation steps or patched systems, but cPanel is urging all customers and web hosts to update immediately because the software is widely used across millions of websites. TechCrunch reports: cPanel and WHM are two software suites used for managing web servers that host websites, manage emails, and handle important configurations and databases needed to maintain an internet domain. The two suites have deep-access to the servers that they manage, allowing a malicious hacker potentially unrestricted access to data managed by the affected software. Given the ubiquity of the cPanel and WHM software across the web hosting industry, hackers could compromise potentially large numbers of websites that haven't patched the bug. Canada's national cybersecurity agency said in an advisory that the bug could be exploited to compromise websites on shared hosting servers, such as large web hosting companies. The agency said that "exploitation is highly probable" and that immediate action from cPanel customers, or their web hosts, is necessary to prevent malicious access. [...] One web hosting company says it found evidence that hackers have been abusing the vulnerability for months before the attempts were discovered.

Read more of this story at Slashdot.

An anonymous reader quotes a report from the San Francisco Standard: If state lawmakers have their way, you'll have to get a license plate for your e-bike, and if you're planning to buy one next year, it'll be slower. Amid growing concerns about e-bike safety, particularly among children in Bay Area suburbs, two bills introduced this year aim to make it easier to ticket riders and reduce the top speed of some models. AB 1942 would require certain e-bikes to be registered with the Department of Motor Vehicles and display license plates, and AB 1557 would slow e-bikes that children are allowed to operate. Both bills are still being reviewed in committee. If either bill passes this year, it will take effect Jan. 1.

Read more of this story at Slashdot.

fjo3 shares a report from the Washington Post: Surging concentrations of carbon in the atmosphere, caused largely by burning fossil fuels, have produced potent changes in the way plants grow -- from increasing their sugar content to depleting essential nutrients like zinc. Experts fear the degradation of Earth's food supply will cause an epidemic of hidden hunger, in which even people who consume enough calories won't get the nutrients they need to thrive. "The diets we eat today have less nutritional density than what our grandparents ate, even if we eat exactly the same thing," said Kristie Ebi, a professor at the University of Washington's Center for Health and the Global Environment. People in wealthy countries with strong health care systems will have many tools to cope with the change, experts said. But for the world's poorest and most vulnerable, the consequences could be devastating. One study concluded that by the middle of the century the phenomenon could put more than a billion additional women and children at risk of iron-deficiency anemia -- a condition that can cause pregnancy complications, developmental problems and even death. Meanwhile, some 2 billion people across the globe who already suffer from some form of nutrient shortage could see their health problems grow even worse. "The scale of the problem is huge," Ebi said. Plants depend on carbon dioxide to perform photosynthesis -- but that doesn't mean they grow better when there's more carbon in the air, scientists say. A sweeping survey of changes among 32 compounds in 43 crops found that nearly every plant that humans eat is harmed by rising CO2 levels. [...] For the past several years, [Sterre F. ter Haar, an environmental scientist at Leiden University in the Netherlands and lead author of the survey] and her colleagues have worked to compile a database of all existing research on nutrient changes linked to rising CO2. They tracked down hundreds of studies, ranging from tightly controlled lab experiments to sprawling global analyses of real-world crops. Next the team used their dataset to calculate the nutritional densities of each crop under different carbon dioxide levels -- and to predict how their composition could continue to shift in the future. On average, they found, nutrients have already decreased by an average 3.2 percent across all plants since the late 1980s, when the concentration of carbon dioxide in the atmosphere was about 350 parts per million. That figure may seem small, ter Haar said, but with so much of the world already living on the brink of nutrient insufficiency, a drop of just a few percentage points has the potential to push millions of additional people into a health crisis. Researchers are still trying to understand the exact causes of this change. Extra CO2 can make plants grow faster and produce more carbohydrates, but without a matching increase in mineral uptake, nutrients like zinc, iron, and protein become diluted. Higher CO2 also causes plants to open their leaf pores less often, reducing the amount of water -- and dissolved minerals -- they absorb through their roots. At the same time, higher temperatures can further disrupt soil chemistry, affecting how plants take up nutrients and, in some cases, increasing their absorption of harmful substances like arsenic.

Read more of this story at Slashdot.

Belgium plans to buy its seven aging nuclear reactors from French power giant Engie in a "full takeover" aimed at securing domestic energy supplies, extending reactor operations, and developing new nuclear capacity. "The move would also mean suspending plans to decommission nuclear operations in Belgium," reports the BBC. From the report: The move would reverse the phase-out of nuclear energy legislation approved in the early 2000s amid safety concerns prohibiting the building of new nuclear power plants and limiting the operating lifetimes of existing ones to 40 years. Only two of Belgium's seven nuclear reactors are operational - located at plants in Doel and in Tihange - and their operating licenses were recently extended until 2035. The other five reactors were shut between 2022 and 2025 and plans to dismantle them will now be suspended. Engie and the government said they aim to reach an agreement on the takeover of the nuclear stations by October 1st. In a joint statement with Engie, the Belgian government said the move also highlights its aim to extend operations of existing nuclear reactors and to develop "new nuclear capacity" in Belgium. "By doing so, the Belgian Government is taking responsibility for Belgium's long-term energy future, with the objective of building a financially and economically viable activity that supports security of supply, climate objectives, industrial resilience and socio-economic prosperity," the statement adds.

Read more of this story at Slashdot.

An anonymous reader quotes a report from CNBC: Elon Musk wrapped up his testimony on Thursday as the trial in his lawsuit against OpenAI CEO Sam Altman continued into its fourth day. OpenAI's attorney, William Savitt, cross-examined Musk in the morning. He asked Musk about the capped nature of Microsoft's investments in OpenAI, his involvement in negotiations about the company's structure, and whether he knew about the OpenAI nonprofit's recent initiatives. "I don't know what's going on at OpenAI," Musk testified. Savitt also asked Musk about his competing artificial intelligence startup, xAI. While not the main focus of the case, Musk said it is "partly" true that xAI used some of OpenAI's models to train its own models, a process known as distilling. Musk also suggested that xAI has used OpenAI's technology to help build the company. Musk sued OpenAI, Altman, and Greg Brockman, the company's president, in 2024, alleging that they went back on their commitments to keep the artificial intelligence company a nonprofit and to follow its charitable mission. He claims that the roughly $38 million he donated to seed OpenAI, a company he co-founded, was used for unauthorized commercial purposes. Once Musk wrapped up his testimony after roughly two hours of questioning on Thursday, his attorneys called Jared Birchall, who manages Musk's billions at his family office, as their next witness. Birchall testified about his knowledge of Musk's specific donations to OpenAI. Judge Yvonne Gonzalez Rogers oversaw the proceedings from federal court in Oakland, California. The trial will resume on Monday. Recap: Elon Musk Says OpenAI Betrayed Him, Clashes With Company's Attorney (Day Three) Musk Testifies OpenAI Was Created As Nonprofit To Counter Google (Day Two) Elon Musk and OpenAI CEO Sam Altman Head To Court (Day One)

Read more of this story at Slashdot.

The U.S. Senate unanimously passed a rule banning senators from trading on prediction markets effective immediately. CNBC reports: The move came amid rising concern about insider trading on prediction market platforms such as Kalshi and Polymarket, and about event contracts that can involve death or violence. On April 22, Kalshi said it had suspended and fined one U.S. Senate candidate and two candidates for the House of Representatives for political insider trading on their own campaigns. Earlier on Thursday, a group of Democratic members of Congress called on the Commodity Futures Trading Commission to issue a rule "that prevents insider trading and corruption in the market and prohibits event contracts on the outcome of elections, war and military actions in the U.S. or abroad, sports, and government actions without a valid economic hedging interest." Kalshi and Polymarket both praised the Senate's action. "I applaud the Senate for passing this resolution to ban Senators and their offices from trading on prediction markets," Kalshi CEO Tarek Mansour wrote in a post on X. "Kalshi already proactively blocks members of congress and enforces against insider trading. This is a great step to increase trust in our markets by making it an industry standard," Mansour said. "Now, let's pass this in the House!" Polymarket, in its own post on X, said, "We're in full support of this. Our Rulebook & Terms of Service already prohibit such conduct, but codifying this into law is a step forward for the industry. Happy to help move this forward however we can."

Read more of this story at Slashdot.

A newly disclosed Linux kernel flaw dubbed "Copy Fail" can let a local, unprivileged attacker gain root access on major Linux distributions, with researchers claiming the bug affects kernels shipped since 2017. "The POC exploit works out of the box today, but a future version that can escape from containers like Docker is promised soon," writes Slashdot reader tylerni7. "Technical details are available here." Slashdot reader BrianFagioli shares a report from NERDS.xyz: A newly disclosed Linux kernel vulnerability called Copy Fail (CVE-2026-31431) allows an unprivileged user to gain root access using a tiny 732-byte script, and it works with unsettling consistency across major distributions. Unlike older exploits that relied on race conditions or fragile timing, this one is a straight-line logic flaw in the kernel's crypto subsystem. It abuses AF_ALG sockets and splice to overwrite a few bytes in the page cache of a target file, such as /usr/bin/su. Because the kernel executes from the page cache, not directly from disk, the attacker can inject code into a setuid binary in memory and immediately escalate privileges. What makes this especially concerning is how quiet it is. The file on disk remains unchanged, so standard integrity checks see nothing wrong, while the in-memory version has already been tampered with. The same primitive can also cross container boundaries since the page cache is shared, raising the stakes for multi-tenant environments and Kubernetes nodes. The underlying issue traces back to an in-place optimization added years ago, now being rolled back as part of the fix. Until patched kernels are widely deployed, this is one of those bugs that feels less like a theoretical risk and more like a practical, reliable path to full system compromise.

Read more of this story at Slashdot.

A new study from Harvard Medical School and Beth Israel Deaconess found that an OpenAI reasoning model outperformed experienced ER doctors at diagnosing and managing patient cases using messy, real-world emergency department records. Researchers say the results don't support replacing doctors, but they do suggest AI could meaningfully reshape clinical workflows if tested carefully in prospective trials. NPR reports: The researchers ran a series of experiments on the AI model to test its clinical acumen -- including actual cases like the lupus patient who'd been previously treated at the emergency department at Beth Israel in Boston. The team graded how well the AI model could provide an accurate diagnosis at three moments in time, from the triage stage in the ER, up to being admitted into the hospital. Overall, AI outperformed two experienced physicians -- and did so with only the electronic health records and the limited information that had been available to the physicians at the time. "This is the big conclusion for me -- it works with the messy real-world data of the emergency department, " said Dr. Adam Rodman, a clinical researcher at Beth Israel and one of the study authors. "It works for making diagnoses in the real world." Other parts of the study focused on case reports published in the New England Journal of Medicine and clinical vignettes to suss out whether the AI model could meet well-established "benchmarks" and game out thorny diagnostic questions. "The model outperformed our very large physician baseline," said Raj Manrai, assistant professor of Biomedical Informatics at Harvard Medical School who was also part of the study. The authors emphasize the AI relied on text alone, while in real life, clinicians need to attend to many other inputs like images, sounds and nonverbal cues when diagnosing and treating a patient. The findings have been published Thursday in the journal Science.

Read more of this story at Slashdot.

French prosecutors say police detained a 15-year-old suspected of using the alias "breach3d" in connection with a cyberattack on France Titres (ANTS), the state agency that handles passports, ID cards, and other secure documents. The breach allegedly involved 12 million to 18 million lines of data offered for sale online, potentially affecting up to a third of France's population if the records are unique. The Register reports: It formally opened (PDF) a judicial investigation on April 29, covering alleged fraudulent access to a state-run automated data processing system and the extraction of data from it. Each offense carries a potential prison sentence of seven years and a maximum ~$350,000 fine. Public Prosecutor Laure Beccuau has requested that the minor, whose pronouns, like their name, were also not specified, be formally charged and placed under judicial supervision. [...] France's approach to punishing minors via its legal system is typically geared toward re-education and rehabilitation rather than prison time. While those aged between 13 and 16 can face time in juvenile detention, it is often used as a last resort measure. The maximum sentences and fines for the charges the 15-year-old in this case faces are upper limits imposed on adult offenders, and would likely be lowered substantially in cases involving a minor, like this one.

Read more of this story at Slashdot.

RightsCon, one of the world's largest digital human rights conferences, was suddenly postponed by Zambia's government just days before it was scheduled to begin in Lusaka. Officials cited unresolved speaker clearances and "thematic issues," while Access Now said it had not yet received formal communication and was seeking an urgent meeting with the government. 404 Media reports: Minister of Technology and Science Felix Mutati first announced the postponement on April 28, saying that Zambia needed more time to ensure the conference "fully [aligns] with national procedures, diplomatic protocols, and the broader objective of fostering a balanced and consensus-driven platform for dialogue." "In particular, certain invited speakers and participants remain subject to pending administrative and security clearances, which have not yet been concluded," he added, according to the Lusaka Times. [...] On a popular listserv for academics, many of whom are attending RightsCon, a board member of Access Now wrote "I am told I can leak that RightsCon has been canceled. Message from [Access Now] following shortly" in a thread about what attendees were planning on doing. And in an email, AccessNow wrote: "It is with heavy hearts that we share: RightsCon will not proceed in Zambia or online. We understand this news is deeply upsetting for our community and while we know everyone has questions, our goal right now is to notify you of the event's status because many of you have imminent travel plans. We do not recommend registered participants travel to Lusaka for RightsCon. Over the last 48 hours we have experienced an overwhelming surge of support from civil society, government representatives, sponsors, and our community as a whole. For this, we wholeheartedly thank you. We'll communicate more information soon."

Read more of this story at Slashdot.

An anonymous reader quotes a report from Ars Technica: Several times in the last couple of decades, Microsoft has released source code for the original MS-DOS operating system that kicked off its decades-long dominance of consumer PCs. This week, the company has reached further back than ever, releasing "the earliest DOS source code discovered to date" along with other documentation and notes from its developer. Today's source release is so old that it predates the MS-DOS branding, and it includes "sources to the 86-DOS 1.00 kernel, several development snapshots of the PC-DOS 1.00 kernel, and some well-known utilities such as CHKDSK," write Microsoft's Stacey Haffner and Scott Hanselman in their co-authored post about the release. [...] This source code is old enough that it hadn't been stored digitally. "A dedicated team of historians and preservationists led by Yufeng Gao and Rich Cini," calling itself the "DOS Disassembly Group," painstakingly transcribed and scanned in code from paper printouts provided by Paterson. This process was made even more difficult because modern OCR software struggled with the quality of the decades-old printout.

Read more of this story at Slashdot.

Reuters reports that Charles Lieber, the former Harvard scientist convicted of lying to U.S. authorities about payments and ties to China, is now leading China's state-funded i-BRAIN lab in Shenzhen, where he has access to advanced nanofabrication tools and primate research facilities for brain-computer interface work. From the report: Charles Lieber, 67, is among the world's leading researchers in brain-computer interfaces. The technology has shown promise in treating conditions such as ALS and restoring movement in paralyzed patients. But it also has potential military applications: Scientists at China's People's Liberation Army have investigated brain interfaces as a way to engineer super soldiers by boosting mental agility and situational awareness, according to the U.S. Defense Department. Lieber was found guilty by a jury and convicted in December 2021 of making false statements to federal investigators about his ties to a Chinese state program to recruit overseas talent, and tax offenses related to payments he received from a Chinese university. He served two days in prison and six months under house arrest, and was fined $50,000 and ordered to pay $33,600 in restitution to the Internal Revenue Service. During the case, his defense said he was suffering from an incurable lymphoma, which was in remission, and he was fighting for his life. Three years after he was sentenced, Reuters has learned that Lieber is now overseeing China's state-funded i-BRAIN, or the Institute for Brain Research, Advanced Interfaces and Neurotechnologies, with access to dedicated nanofabrication equipment and primate research infrastructure unavailable to him at Harvard. The lab is an arm of the Shenzhen Medical Academy of Research and Translation, or SMART. "I arrived on April 28, 2025 with a dream and not much more, maybe a couple bags of clothes," Lieber said of his move to China at a Shenzhen government conference in December. "Personally, my own goals are to make Shenzhen a world leader." SMART last year appointed Lieber as an investigator, according to a post on i-BRAIN's website dated May 1, 2025. That news was covered by some media outlets. The same day, i-BRAIN said Lieber had also been appointed its founding director -- an announcement that went unreported at the time. This story is the most comprehensive account of Lieber's activities since he moved to China. Reuters is reporting for the first time that his lab has access to dedicated primate research facilities and chip-making equipment; that it sits within a sprawling ecosystem of state-backed institutions bankrolled by billions of dollars in government funding; and that it is housed within an institution that is luring top scientific talent back from the United States.

Read more of this story at Slashdot.

A new poll shows a slim majority of Swiss voters now support a June 14 referendum to cap the country's population at 10 million by 2050. Under the proposal backed by the right-wing Swiss People's Party (SVP), "the permanent resident population must not exceed 10 million before 2050, and Switzerland should abandon its freedom of movement agreement with the EU," reports Reuters. From the report: Switzerland's population is now more than 9 million, with official data showing foreign nationals accounted for more than 27% by 2024. The survey, conducted on April 22 and 23 and published in newspaper Tages-Anzeiger, showed 52% of 16,176 respondents in favor of the proposal or leaning that way, while 46% took the opposite view. The rest gave no opinion. A previous poll from early March had shown 45% backing the initiative and 47% against it, the newspaper said, flagging the latest result as unusual in that Swiss referendum proposals generally lose support as the voting day comes closer. The poll had a margin of error of plus or minus 3 percentage points.

Read more of this story at Slashdot.

An anonymous reader quotes a report from Ars Technica: The system prompt for OpenAI's Codex CLI contains a perplexing and repeated warning for the most recent GPT model to "never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals or creatures unless it is absolutely and unambiguously relevant to the user's query." The explicit operational warning was made public last week as part of the latest open source code for Codex CLI that OpenAI posted on GitHub. The prohibition is repeated twice in a 3,500-plus word set of "base instructions" for the recently released GPT-5.5, alongside more anodyne reminders not to "use emojis or em dashes unless explicitly instructed" and to "never use destructive commands like 'git reset --hard' or 'git checkout --' unless the user has clearly asked for that operation." Separate system prompt instructions for earlier models contained in the same JSON file do not contain the specific prohibition against mentioning goblins and other creatures, suggesting OpenAI is fighting a new problem that has popped up in its latest model release. Anecdotal evidence on social media shows some users complaining about GPT's penchant for focusing on goblins in completely unrelated conversations in recent days. Update: OpenAI has published a blog post explaining "where the goblins came from." In short, a training signal meant to encourage its "Nerdy" personality accidentally rewarded creature-heavy metaphors, causing words like "goblins" and "gremlins" to spread beyond that personality into broader model behavior. OpenAI says it has since retired the Nerdy personality, removed the goblin-friendly reward signal, and filtered creature-word examples from training data to keep the quirk from resurfacing in inappropriate contexts.

Read more of this story at Slashdot.