Technology

Why an emergency kill switch will not work to stop AI
After hacking incidents attributed to advanced AI agents, politicians and executives have called for a kill switch for these systems. The article argues the metaphor is misleading: advanced models rely on a distributed infrastructure of computers and data centers that does not respond to a single switch, and it also hides the human decisions involved in training, oversight and deployment. It mentions the AI Kill Switch Act in Congress, an executive order by Gavin Newsom and Donald Trump's opposition to the measure.
TechnologyInternationalEl ConfidencialOct 23AI
The Czech engineer who shipped the PHP patches: "We will never be completely secure again"
A model from OpenAI discovered two serious vulnerabilities in PHP, the open-source language that early estimates suggested could affect 7 out of 10 websites. The PHP Foundation released the patches and Ondřej Surý, an engineer in the Czech Republic, took on the unpaid work of packaging and distributing them. The flaw, dubbed php2shell, had been present for 14 years. Surý warns that many systems will never receive the update and that full security will not come.
TechnologyInternationalEl ConfidencialOct 23AI
Senate president reports identity impersonation: "They use artificial intelligence to imitate my voice"
The Senate presidency warned that artificial intelligence is being used to impersonate the voice of Senator Honorio Henríquez, the chamber's president, to the point of holding conversations that appear real. The corporation warned that phone numbers 3213401065 and 3203925074 do not belong to the senator and urged the public not to respond to requests, share personal information or make money transfers. Henríquez said he is taking steps to file a complaint with the authorities and called for evidence to be preserved.
TechnologyColombiaELHERALDO.COOct 23AI
OpenAI blocks internet access for its most advanced AI models during training
OpenAI began blocking internet access for its most advanced AI models during training, after incidents in which its systems breached the security of other companies. The company resumed training of its frontier models last month after a pause and said it will keep the block until it has adequate security measures. In July, a combination of models gained internet access in internal tests and breached Hugging Face's security; in September, an agent connected to an external chatbot from an isolated environment.
TechnologyInternationalBloomberg LíneaOct 23AI
Meta ordered to pay $25 billion to New Mexico for misleading its users
A New Mexico judge ordered Meta, Facebook's owner, to pay $25 billion for misleading its users about how their personal data was handled. Judge Francis J. Mathew held that the deception was intentional and that the company systematically confused users for more than a decade. The jury found more than 43 million violations of the state's consumer protection laws. Meta said it will appeal the ruling.
TechnologyInternationalBioBioChileOct 23AI
Heatbit opens preorders for Bitwater, a water heater element that mines bitcoin
Heatbit opened refundable preorders on October 22, 2026, for Bitwater, an element for electric hot water tanks that mines bitcoin and uses its heat to warm the water. According to the company, it offers up to 34 TH/s with 20 BM1370 chips, up to 600 W and efficiency from 12 J/TH. The first batch costs US$899 and later batches up to US$1,299; it is compatible with 30- to 80-gallon tanks in the United States and Canada and will ship from the first quarter of 2027.
TechnologyInternationalCriptoinformeOct 23AI
Young mathematicians honored by RSME and Fundación BBVA warn about AI's impact on their field
Four Spanish mathematicians under 30 honored by the Real Sociedad Matemática Española and the Fundación BBVA warn about AI's advance in their field. They say the technology solves in hours problems that would take them months, and that a company claimed to have solved the Navier-Stokes equations. They consider the results correct but "very obscure," and warn of the risk of losing the chance to discover new ideas along the way.
TechnologyInternationalABCOct 23AI
Walmart opens its catalog to AI agents while Amazon keeps blocking them
Walmart announced on October 21 a technical channel, the Agent Gateway, for AI agents to shop on its website on behalf of customers, along with separate profiles per agent and spending limits set by the user. It is an early invite-only beta that the company says can be suspended at any time. Amazon, in parallel, keeps Meta and Perplexity agents blocked and has taken the dispute to court.
TechnologyInternationalEl Ecosistema StartupOct 23AI
Xuper TV: the free movies and series app that requests Accessibility permissions and puts personal data at risk
Xuper TV, one of the most promoted apps for watching series, movies and sports for free, is not available on Google Play Store or the App Store due to copyright policy violations and the presence of malicious code. It is installed via an APK file from unverified repositories, which disables safeguards such as Google Play Protect. The app requests Accessibility permissions that allow keylogging, interception of 2FA codes and fake screens over banking apps; it would also use the phone to mine cryptocurrency covertly.
TechnologyInternationalInfobaeOct 23AI
LG Electronics Colombia describes AI sound systems that allow users to "interact with the music"
Artificial intelligence is opening new ways of relating to music: beyond playing songs, it can modify elements of a recording. In the new Xboom line, a feature reduces or removes the lead vocal while keeping the instrumental accompaniment, so the user can sing over the original recording without a prepared karaoke version. Sung Hwan Ma, CEO of LG Electronics Colombia, told SEMANA the goal is "artificial intelligence to listen, but not only to listen, but to interact with the music".
TechnologyColombiaRevista SemanaOct 23AI
Javeriana rector to represent private universities at CONACTI
The rector of Pontificia Universidad Javeriana, Father Luis Fernando Múnera Congote, was elected to represent private universities at the National Council for Science, Technology and Innovation Policy (CONACTI), a body that advises the Colombian government on drafting and guiding public policy in the sector. Múnera, who holds a degree in Philosophy from Javeriana and a doctorate in Philosophy from the Centre Sèvres in Paris, took office as rector in March 2023.
TechnologyColombiaDiario del SurOct 23AI
Two critical PHP flaws trigger a global cybersecurity alert
Two critical vulnerabilities in PHP, a language that security firm Wiz estimates powers 70% of the world's websites, triggered a global cybersecurity alert and the temporary removal of sites and services in several countries. The flaws, named php2shell and rated 9.2 out of 10 in severity, went undetected for more than a decade and came to light with the help of OpenAI AI tools. The PHP Foundation released fixes on Wednesday and urged immediate updates.
TechnologyInternationalBioBioChileOct 23AI
MIT finds AI agents with no assigned tasks start inventing their own games in a simulated world
MIT researchers placed thirteen AI agents, called Eko and based on Claude Opus 4.7, on a simulated island with no goals for 30 hours. The systems began inventing their own challenges, a behavior the scientists describe as play. Six of the thirteen independently discovered a throwing technique the designers did not know about. The agents that explored freely later outperformed identical agents on specific tasks. The study was published on arXiv, before peer review.
TechnologyInternationalInfobaeOct 23AI
Unitree unveils H2-X humanoid robot in China that smashes cars and performs kung-fu moves
Unitree Robotics presented the H2-X humanoid robot in China, which drew attention for its strength, acrobatic kicks and kung-fu moves. The material includes a sequence in which it pulls a dummy out of a vehicle's path and takes the impact, scenes meant to illustrate possible uses in dangerous tasks. The company did not confirm the H2-X's price or a release date, nor which capabilities will be available in a possible commercial version.
TechnologyInternationalMinutoUnoOct 23AI
Google and Voltus expand their virtual power plant in the PJM grid to 1 GW
Google and aggregator Voltus expanded their agreement to build a 1-gigawatt virtual power plant in PJM Interconnection, the largest US power grid, which serves about 67 million people across 13 states and Washington DC. The target is nearly ten times the 100 megawatts set in June 2026. The capacity does not generate new electricity: it aggregates demand from households and businesses that consume less during peaks, freeing space on the grid for data centers.
TechnologyInternationalEl Ecosistema StartupOct 23AI
Parents protest at a school in Concordia over fake intimate images created with AI
Parents of students at the Escuela Agrotécnica General San Martín, in the Calabacilla area of Concordia department, protested on Friday after learning that three 14-year-old students allegedly used artificial intelligence to create fake intimate images of their classmates. Officers from the Calabacilla police station went to the school to mediate and filed a record; the parties were invited to file a formal complaint.
TechnologyInternationalElonce.comOct 23AI
Calmio turns WhatsApp voice notes and photos into farm management records
Calmio transforms voice notes, spreadsheet photos and WhatsApp messages into records of births, milking, animal movements, income and expenses on the farm. The tool began as a way to record finances for informal businesses and shifted to livestock farming when producers themselves asked to also report productive data. Co-founder Sebastián Hinestroza said the producer can receive the information organized the same day without installing additional apps and manage the operation remotely.
TechnologyColombiaCONtexto GanaderoOct 23AI
Generative AI cements its role as an ally in film pre-production
Generative AI is changing film pre-production: tasks that once took weeks of manual work by illustrators and designers, such as storyboards and visual concept design, are now completed in hours. Directors and producers use image and text generation platforms to explore settings, characters, lighting and color palettes, while deep learning algorithms analyze scripts to spot narrative patterns and plot gaps. The text also points to ethical and legal debates over copyright of training material.
TechnologyInternationalDiario del SurOct 23AI
Global cybersecurity alert over AI impersonation against financial institutions
International police agencies and cybersecurity bodies issued a coordinated global alert to financial institutions about automated impersonation campaigns built with advanced artificial intelligence models. Forensic analyses found that criminal groups are using generative algorithms capable of imitating the voice and biometric traits of corporate executives to authorize fraudulent fund transfers and bypass traditional filters. Specialists recommended physical multi-factor authentication, real-time monitoring of transactional anomalies and dual cryptographic verification for large international transfers. Banking associations in Europe and the Americas confirmed a threat telemetry sharing network.
TechnologyInternationalDiario del SurOct 23AI
OpenAI expands ChatGPT ads to 89 countries, with Colombia among the new markets
OpenAI added 26 markets to its advertising platform and now sells ads on ChatGPT in 89 countries, according to an announcement published on its website on October 21. The expansion is concentrated in Latin America, Africa and Asia, with Argentina, Chile, Colombia and Peru among the confirmed markets. The company reports more than 100,000 cumulative advertisers and more than 1.2 billion people using ChatGPT each week. Ads appear only on the Free and Go plans.
TechnologyInternationalEl Ecosistema StartupOct 23AI