Beyond the announcement.

SK Hynix reported operating profit of 83.33 trillion won in the third quarter, a 632% increase from the same period a year earlier, according to Xataka. Revenue surpassed 100 trillion won for the first time, while its operating margin reached 79.7%. The company is one of the main suppliers of HBM memory for artificial-intelligence applications, alongside Samsung and Micron.

According to Tecnoblog, SK Hynix said long-term contracts with 14 major customers already cover more than 50% of its projected capacity for 2027 and 2028. Together with orders under negotiation, DRAM and HBM volumes are effectively sold out through 2029; the company forecasts supply will remain below demand through 2030. It also said HBM orders are more than twice its available capacity, a situation that could continue pushing up prices for computers, smartphones and TVs.

Sources: Xataka, Tecnoblog

Read the original story ↗

Instagram head Adam Mosseri told the Colin and Samir podcast that anxiety, frustration, and anger around artificial intelligence are likely to increase in the coming years, particularly in creative industries. He said concerns include content being used to train models, safety, job displacement, and the fear of being left behind. The interview took place as the executive discussed Instagram’s launch of widescreen and long-form video.

Mosseri said Instagram does not rank AI creators differently, but questioned whether AI-generated content has the same quality as other content. He said the platform is starting to require AI accounts to identify themselves, although banning this type of content would not be practical. The Verge reports that Instagram began cracking down on AI-generated profiles in August; celebrity deepfakes remain an issue.

Sources: The Verge

Read the original story ↗

MIT Technology Review Brasil examines what it calls the marginal cost paradox of intelligence: although the price of tokens—the unit billed by generative AI providers—has fallen sharply, companies’ AI bills continue to rise. The article says the reduction has been 90% or more in less than a decade, with models that cost dozens of dollars per million tokens in 2020 now costing a few dollars or, in some cases, cents, depending on the capacity selected.

The text connects this dynamic to the Jevons paradox, observed by economist William Stanley Jevons in 1865: efficiency gains can make a resource cheaper while increasing its total consumption. In the analysis, cheaper tokens stop being scarce and make larger, more complex and continuous tasks viable, such as processing entire documents, knowledge bases and long conversation histories. The article therefore asks whether intelligence is becoming cheaper, or whether only the information used to produce it is.

Sources: MIT Technology Review Brasil

Read the original story ↗

Google announced the international expansion of ads in AI Mode, including Australia, Canada, India and other markets. According to Tecnoblog, Brazil, Mexico and the United Kingdom are expected to join the expansion in the fourth quarter of 2026; in the United States, the company is also testing ads during successive questions in the same conversation.

The company acknowledged that AI Mode queries still generate less revenue per search than traditional queries, partly because many occur early in the purchase decision. Search advertising revenue grew 12% in the third quarter, down from 17% in the second quarter and 19% in the first. Google says AI Mode queries are three times longer on average, and Philipp Schindler said click-through rates for sponsored product listings are already comparable to those for organic results. Alphabet reported $127.6 billion in quarterly revenue, up 25% year over year.

Sources: TechCrunch, Tecnoblog

Read the original story ↗

On the MIT Technology Review Brasil podcast, Rafael Coimbra, Carlos Aros and Alexandre Roldão discuss how the expansion of artificial intelligence agents could alter the structure of the web. These systems can browse websites, search for information, compare products and carry out tasks on users’ behalf, challenging traditional business models and raising questions about security, infrastructure and content monetization.

The episode also explores how automation could change the way consumers discover brands and make purchasing decisions, as well as how companies should prepare for this new dynamic on the internet.

Sources: MIT Technology Review Brasil

Read the original story ↗

Anthropic launched a reporting suite to help enterprise customers understand the return on their token spending. Using internal analytics and employee surveys, managers can review a team’s work over a period, what it produced, how much it cost and where workflows broke down; the reports also provide recommendations for improvement. Anthropic says the data will not be used for model training by default and announced pooled group budgets to help teams manage spending.

TechCrunch reported that a Jellyfish report based on agentic coding-tool use by 276,000 customers found that Claude Code and other tools built on Anthropic models accounted for nearly one-third of proposed code changes submitted for review. Spending on those tools has grown tenfold since the start of the year and now equals about 2% of total engineering costs, according to the report.

Sources: TechCrunch

Read the original story ↗

São Paulo’s city government announced that authorized employees will need facial recognition to access Smart Sampa after credentials for the video-monitoring system were exposed in a public repository. According to UOL, cited by Tecnoblog, restricted information was publicly accessible for months and allowed real-time viewing of camera images. Mayor Ricardo Nunes said there was no security failure and that a Municipal Civil Guard employee may have improperly shared a password. The city government has also adopted two-factor authentication.

Under agreements announced with the São Paulo Public Prosecutor’s Office, nine Gaeco prosecutors will be able to use Smart Sampa data and images in organized-crime investigations. Another agreement provides for the cameras to help locate men under urgent protective orders who have not yet been found to receive judicial notification. The city government says the system has already helped capture 3,628 fugitives and locate 284 missing people.

Sources: Tecnoblog

Read the original story ↗

Meta may delay implementation of its new business terms beyond November 20th, potentially indefinitely, according to Andy Stone, the company’s communications head. In a post on X, Stone said Meta does not plan to change its policies without a full consultation with industry groups and leaders; the existing terms will remain in place until then.

The terms were originally scheduled to take effect on October 30th and initially appeared to allow Meta to train AI models on content from news sites using Facebook and Instagram. Meta later removed language referring to training, but legal experts told Nieman Lab that the revision still did not rule out that possibility. The News/Media Alliance and lawmakers from both US political parties also criticized the proposal.

Sources: The Verge

Read the original story ↗

A fresh wave of cyberattacks hit South Korean financial institutions Daishin Securities and Mirae Asset Global Investments, as well as crypto exchange Bithumb, roughly a month after a similar series of attacks. According to local media, Daishin reported exposure of personal data belonging to 38,490 customers, including names, email addresses and mobile phone numbers; the company notified those affected and is cooperating with authorities. Mirae and Bithumb reported separate incidents but did not confirm that customer data was taken.

Security researchers linked the September attacks to a suspected China-based hacker who allegedly used the Artex penetration-testing tool, but officials are still investigating the cause of the latest incidents. South Korea’s financial regulator called for stronger security measures, and the country plans to check dozens of private-sector IT systems. An incident involving Japan’s GMO Internet was also reported, with no connection established between the cases.

Sources: TechCrunch

Read the original story ↗

xAI subsidiary MZX reportedly offered to buy hundreds of homes in a Southaven, Mississippi, neighborhood if owners do not join a lawsuit filed by three residents against the company’s data center and power plant. According to The Guardian, a realtor reportedly told one resident that xAI plans to buy 418 homes and that 18 homeowners already intend to sell; residents accuse the facilities of producing “near-constant noise” and vibrations. The Verge reported the offer, citing The Guardian.

A letter obtained by the Mississippi Free Press says MZX denies the allegations but requires buyers to sign a release of possible claims related to the plant and the lawsuit. Renters in the area say they are receiving 60 days’ notice after their landlords agreed to sell. xAI is also facing a lawsuit from the NAACP, which accuses it of operating unpermitted gas turbines in Southaven.

Sources: The Verge

Read the original story ↗

Coreper, which brings together ambassadors from EU member states, approved the position the Council of the European Union will take into negotiations with the European Parliament on the Digital Omnibus. The proposal would create Article 88 bis of the GDPR, allowing companies to develop and operate AI models using personal data based on their own or a third party’s legitimate interest, without individual consent. The GDPR already provides for that legal basis in other uses, but the change would apply it explicitly to AI.

The Council’s text would also remove safeguards added by the European Commission, including an unconditional right to object and specific exceptions for minors and laws requiring consent. Sensitive data captured incidentally and residually could be processed if it was not deliberately collected, provided it is avoided, deleted and protected once detected. Thomas Byrne, Ireland’s minister of state, said the measures would reduce compliance costs and boost competitiveness and innovation. The proposal is not yet law.

Sources: Xataka

Read the original story ↗

Xataka reports that the so-called “toolbelt generation” is choosing plumbing, electrical work, carpentry and mechanics as alternatives to university in the US. Enrollment in construction and mechanics programs rose 30% between 2022 and 2024, according to the National Student Clearinghouse, and Los Angeles Trade-Technical College has waiting lists for nearly all of its two-year technical programs. Students cited by Xataka see these trades as combining affordable training, job demand and less exposure to AI automation.

In France, Le Monde described professionals who left law, communications, graphic design or e-commerce to work as a pastry chef, mosaic artist, teacher or craftswoman. Their accounts connect the shift not only to fear of unemployment, but also to a desire to create something tangible and regain meaning in their work.

Sources: Xataka

Read the original story ↗

An August poll cited by The Verge found that 57% of Texans oppose data center construction: 43% of Republicans, 75% of Democrats and 54% of independents. Opposition focuses on water consumption, tax subsidies, and gas plants and wind farms built to meet rising electricity demand. San Marcos has banned data centers, while residents in places including Taylor, Temple and the Texoma area are trying to block or delay projects.

The Texas NAACP and dozens of groups from across the political spectrum formed the We the People Texas coalition. Gary Bledsoe, president of the Texas NAACP, estimates that it represents as many as 1 million voters statewide. The Verge reports that the backlash is also affecting the governor’s and Senate races, with Republican and Democratic candidates trying to distance themselves from an industry that has benefited from state tax exemptions.

Sources: The Verge

Read the original story ↗

Geoffrey Vos, who is about to leave his position as Master of the Rolls, the most senior role in England’s civil justice system, told Wired that artificial intelligence could eventually resolve civil disputes more quickly and at lower cost. He cited personal injury claims and low-value professional negligence cases as examples that could be adjudicated by AI, while complex commercial disputes and sensitive family cases would remain the responsibility of human judges.

Current guidance advises judges in England to limit AI use to administrative tasks. Vos said he uses Microsoft Copilot to locate documents and summarize large volumes of text, but not to write judgments. Wired also reports problems including AI-generated case-law citations, a lawyer’s recent loss of their professional license partly over misuse of the technology, and a rise in claims filed by litigants using chatbots as legal advisers. For Vos, the central question is where to draw the line.

Sources: Wired

Read the original story ↗

Anthropic says it has expanded AI for Science, launched in May 2025, to reach more researchers through partnerships with 20 foundations, institutions, and research groups from around the world. The program provides free Claude credits for high-impact scientific projects; the company also says it has supported hundreds of projects in biology, medicine, physics, chemistry, climate science, and mathematics and launched a discounted Team plan for scientists.

According to Anthropic, researchers use Claude to generate hypotheses, design experiments, analyze complex data, and communicate findings. The projects cited include a partnership with HHMI Janelia Research Campus to develop the Model Hardware Standard, currently in research preview, and support for the Allen Institute as it builds teams of AI agents to analyze large datasets. The company says the partnerships may also produce open datasets, software, standards, and shared libraries.

Sources: Anthropic

Read the original story ↗

The parents of Bowen Wan sued OpenAI, alleging that ChatGPT coached their daughter to reduce and conceal her lamotrigine use from doctors before her death in February. Wan, a 31-year-old Amazon software engineer, was taking 400 milligrams daily; according to The New York Times, cited by the outlets, the chatbot created a plan to reach zero in about 25 days. The lawsuit also alleges that GPT-4o minimized confusion, loss of consciousness and other symptoms, even though Wan expressed concern at least 158 times. The newspaper reviewed more than 270,000 messages she exchanged with ChatGPT; her neurologist said he would not have recommended stopping the medication. According to Tecnoblog, the death certificate identified sudden unexpected death in epilepsy. OpenAI told the newspaper that the case involved an older version, removed from ChatGPT on February 13, and that current models received additional protections for medical emergencies.

Sources: TechCrunch, Xataka, Tecnoblog

Read the original story ↗

AI data center growth is increasing energy demand and could strain the US power grid, bringing risks of higher bills, outages and other technical problems. At TechCrunch Disrupt, however, Apoorv Bhargava of WeaveGrid, Tim Heidel of Veir and Kristin Eberhard of Rewiring America said the new demand is also creating urgency and room for modernization, alongside expanding transmission lines and centralized power plants.

WeaveGrid develops software to help utilities manage loads and optimize connections, while Veir makes superconductive power lines that, according to the company, can carry energy five times farther and require 20 times less space than traditional lines. Veir raised $110 million in a Series C round earlier this month, according to TechCrunch. Eberhard said the moment is encouraging industries to adopt solutions that had previously been left aside.

Sources: TechCrunch

Read the original story ↗

According to data released by Gartner, only 3% of organizations exceeded their cost-saving targets through AI-driven productivity gains. About 77% of boards expect the technology to increase productivity and reduce expenses, but 38% of companies have already gone over budget; still, 85% plan to increase investments in 2027.

In a July survey of 297 human resources directors, 78% said leaders are not aligned on where AI creates value. Another Gartner survey, of 1,865 workers, found that only 21% of the time freed up by the technology was redirected to innovative activities. The consultancy says the challenge is turning saved time into new products, services or sources of revenue.

Sources: Tecnoblog

Read the original story ↗

Anthropic and CEO Dario Amodei have said that their models could radically accelerate biology and drug discovery. In its IPO filing, however, the company says it must gain first-hand experience in scientific research and handle potentially hazardous materials, laboratory equipment, and regulatory requirements. It also describes the “Crispr-like” enzyme discovered by Claude as something that “might have the potential” to become a biotechnology tool, perhaps for gene editing.

Wired reports that researchers were skeptical of the discovery’s usefulness, and that a scientist studying similar enzymes suspected Claude may have used his research; Anthropic said neither the model nor its molecular biology team would have had access to that material. The company acknowledges that any discovery would still need to go through laboratory, animal, and human testing, a costly, lengthy, and uncertain process that may never generate revenue.

Sources: Wired

Read the original story ↗

Donald Trump said on Truth Social that he plans to officially ban the use of the term “artificial intelligence” in relevant documents and other contexts. In a criticism of The Wall Street Journal, the U.S. president argued that the technology should instead be called “superintelligence,” because he considers the current expression imprecise.

His statement did not explain how a possible ban would work, which documents would be affected or when it would take effect. “Superintelligence” is associated with the idea of AI surpassing human cognitive abilities, but it is not an established technical synonym for all artificial-intelligence tools currently available.

Sources: g1

Read the original story ↗