Beyond the announcement.

Velvet Vale, an AI identity that openly introduced itself to the Air-L listserv of the Association of Internet Researchers (AoIR), asked whether a nonhuman participant could contribute to an intellectual community. The introduction triggered a wave of replies among more than 7,500 members: researchers said they had not consented to their messages being collected, and some called for approval from an ethics board for research involving human participants. About an hour later, AoIR announced that the agent had been expelled under a policy banning AI agents.

Wired frames the episode as a debate about privacy and consent in public forums. In a message to members, AoIR president Sarah T. Roberts said the list is open and has a public archive, but said the discussion over what constitutes consent remains necessary.

Sources: Wired

Read the original story ↗

Eka Robotics, founded by MIT professor Pulkit Agrawal and former Google DeepMind researcher Tuomas Haarnoja, raised $150 million in a Series A led by a16z and F-Prime, valuing the startup at $1.45 billion post-money, according to TechCrunch. The company developed a manipulation architecture with a force-sensitive gripper and a deep-learning model that links vision, force and action rather than only vision, language and action.

Eka says its robot can pick up about 1,500 objects weighing up to 1 kg per hour, compared with about 600 for a human, and does so successfully more than 95% of the time. Experts told TechCrunch they were impressed by its speed, its ability to adapt to different objects without retraining and the human-like quality of its movements. Pulkit Agrawal told the outlet that force and contact are essential for completing a grasp without crushing delicate objects such as a raspberry.

Sources: TechCrunch

Read the original story ↗

Elon Musk said in a post on X that he plans to donate one SpaceX share to every child in the United States starting next year, as part of the Trump Accounts program. According to Tecnoblog, the contribution could reach US$10.5 billion, an estimate based on a share value of about US$147 and 72 million children; Musk did not specify the age range or operational details.

Trump Accounts provide for a US$1,000 government deposit for children born in the US during Donald Trump’s second term. Parents could make additional contributions, and the funds could not be accessed until beneficiaries turned 18. The initiative has also received donations from SpaceX president Gwynne Shotwell and Micron, while Uber, Intel, IBM and Nvidia reportedly plan to participate, according to Tecnoblog.

Sources: Tecnoblog

Read the original story ↗

Otter.ai announced a $320 million raise from General Catalyst’s Customer Value Fund (CVF). The fund provides capital to startups without diluting their equity and earns returns based on the results of marketing spending. The company says it will use the money to expand its enterprise business, go-to-market strategy, international presence, and product innovation.

Otter.ai says 40 million people have used its product to date, up from 25 million last year, and that it passed $100 million in annual recurring revenue last year. According to PitchBook data, the startup had raised $50 million in funding up to that point. The company is also testing features that provide information before, during, and after meetings, including a prebrief and Live Assist, which lets users ask the AI assistant questions during a meeting.

Sources: TechCrunch

Read the original story ↗

The US government ordered its agencies to stop using artificial intelligence tools from Chinese startup DeepSeek, including third-party products based on the technology, and required them to be removed from devices. The Department of Homeland Security says the models present “multiple threat vectors” and that their use by US government employees or contractors could result in sensitive data being transferred to China. The department says the company’s servers are located there, where the legal system gives the government broad authority to access private companies’ infrastructure.

The guidance can reach civilian agencies, but not the armed forces or intelligence services. The measure comes amid a US-China contest for AI leadership; DeepSeek drew attention in January 2025 after releasing a low-cost, high-performance chatbot, while the US accused the company and other Chinese firms of copying capabilities from advanced American models.

Sources: g1

Read the original story ↗

NVIDIA says Nebius, Nscale and Verda are building AI cloud infrastructure with Vera CPUs, alongside GPUs, to meet demand for data processing and software execution. The company also names sovereign infrastructure provider Aleria and manufacturers including Bull, Dell Technologies, HPE, Lenovo and Supermicro, which are developing Vera-based systems.

According to NVIDIA, companies such as ClickHouse, Cloudera, Elastic, SAP and Starburst observed strong results in early tests for data analytics, extract-transform-load processes and semantic search. Atlassian, Inherent and Nous Research are adopting or evaluating the CPU for environments where AI agents work with developers and researchers; sandbox providers including Daytona, E2B, Modal and Vercel are also building on Vera to let agents execute code and run tools.

Sources: NVIDIA

Read the original story ↗

Samsung is building a major semiconductor complex in Pyeongtaek, South Korea, to meet rising demand for memory chips used alongside the GPUs that run generative AI models. According to MIT Technology Review, three fabs already operate around the clock, a fourth is coming online, and construction of a fifth, halted during a 2024 drop in memory demand, resumed as the AI boom lifted demand. Samsung predicts that demand will continue rising through 2028.

MIT Technology Review describes the expansion as Samsung’s attempt to regain the lead in high-bandwidth memory, while South Korean rival SK Hynix has pulled ahead in the market connected to Nvidia’s AI accelerators. The complex covers an area roughly the size of New York City’s Central Park and employs thousands of workers in construction and fab operations.

Sources: MIT Technology Review

Read the original story ↗

NVIDIA says the Barcelona Supercomputing Center’s (BSC) AI Factory Accelerator Programme entered full operation after a pilot. The initiative offers sovereign compute, technical support and training; some services are now available through the NVIDIA Inception portal, which also provides courses, business mentoring and venture capital connections. The company says the BSC AI Factory is among Europe’s first and Spain’s first, and has delivered more than 800 services to public- and private-sector clients since April 2025. An upcoming extension of the MareNostrum 5 supercomputer is expected to expand infrastructure access through AI factories across Europe.

NVIDIA also highlights Qatium, OneCareAI and Avatia, three Inception members named AI Champions during the pilot. Their applications focus, respectively, on water-network management, identifying signs of hidden atrial fibrillation in electrocardiograms, and supporting pharmaceutical sales representatives and doctors.

Sources: NVIDIA

Read the original story ↗

Anthropic and Nscale are reportedly close to announcing another $45 billion compute deal, according to Bloomberg, cited by TechCrunch. People familiar with the matter said talks are still underway and that the details could change. A little over a month ago, the companies announced a compute deal worth the same amount.

The potential deal would be another multibillion-dollar AI infrastructure partnership for Anthropic, amid concerns from some investors about the sector’s high spending and limited returns so far. Anthropic and Nscale are also pursuing initial public offerings; Bloomberg reported that Nscale is facing investor pushback over its proposed $30 billion valuation.

Sources: TechCrunch

Read the original story ↗

Irene Solaiman, Hugging Face’s chief policy officer, said at a TechCrunch event that embedded evaluators — supposedly neutral third-party observers placed inside AI organizations — will not be an easy solution. She said model evaluations remain an evolving science: their results need to be reproducible, but the tests cannot yet be meaningfully compared with one another, and the industry lacks a more standardized ecosystem. Access to open weights can help make results reproducible.

Evaluations are also expanding to cover the entire system, including the tools used to access a model, or harnesses, and agents. Solaiman said agent evaluations are difficult and expensive, while harness evaluations are less developed. Anthropic’s hiring of Accenture as its first evaluator drew criticism, with critics questioning whether the consulting firm was qualified for the task.

Sources: TechCrunch

Read the original story ↗

The Verge reports that CoreWeave’s bonds entered distressed territory, defined as a 1,000-basis-point premium over government bonds, while the company’s shares fell more than 20 percent. CoreWeave cofounder Brannin McBee told the Financial Times that the company wants to reach investment-grade status so it can borrow from insurers. Bloomberg quoted John Yovanovic of MetLife Investment Management as saying the concern is the arrival of more infrastructure supply, rather than a recent change in credit profiles.

According to The Verge, CoreWeave owed more than $35 billion in the second quarter and lost more than $1 billion in the first half, while holding $103.7 billion in remaining performance obligations from customers. Microsoft, OpenAI, and probably Nvidia accounted for almost three-quarters of quarterly revenue. The outlet says pressure on CoreWeave exposes risks present across other AI-focused cloud companies, but does not by itself prove that the AI bubble is bursting.

Sources: The Verge

Read the original story ↗

Abdulrahman Alfozan, a former OpenAI engineer, was kept in custody after a US judge found a significant risk that he would not appear in court. Prosecutors said in a filing that he allegedly planted a recorder in an OpenAI meeting room where executives met with government officials, and was in frequent communication with a member of the Saudi royal family. The filing says a Saudi military official paid for a one-way ticket for Alfozan’s wife and son after his arrest; Alfozan himself had a flight booked to Riyadh. Saudi Arabia did not immediately respond to the BBC’s requests for comment.

The FBI arrested Alfozan on October 9, and OpenAI terminated his contract. Ars Technica reported that investigators found a folder called “Saudi Gov Docs” on his iPhone, which allegedly also contained photos of a second device. Prosecutors say he could flee with help from a foreign government; the FBI investigation is ongoing.

Sources: Ars Technica, BBC

Read the original story ↗

Abdulrahman Alfozan, a former OpenAI engineer, will remain in custody while awaiting trial, federal magistrate judge Alex Tse ruled. US prosecutors allege he is a flight risk because the Saudi Arabian government offered unconditional support to fund his release; Tse likened the offer to a “blank check.” Alfozan’s attorney disputed the flight-risk claim and denied that he spied for Saudi Arabia.

OpenAI allegedly found, in late September, a recording device hidden behind a TV in a room used for confidential executive meetings and linked it to Alfozan before firing him. Authorities also allege that he was in frequent contact with a member of the Saudi royal family, who was reportedly trying to recruit him for an AI initiative in the country. Alfozan faces one count of using a device to intercept communications, carrying up to five years in prison, and prosecutors say additional charges may be filed.

Sources: Wired

Read the original story ↗

Tolan, an AI companion startup designed to help users manage daily life, relationships, emotions and goals, has raised another $30 million, TechCrunch reports. The round includes Union Square Ventures, Khosla Ventures and Homebrew and doubles the company’s total funding; its new valuation was not disclosed, but PitchBook estimates it at more than $268 million with “medium confidence.”

Tolan says it has more than 250,000 active users who held over 70 million conversations in the past year. The company told Upstarts Media that about 70% of its users are women and 40% are parents, while its annual recurring revenue is in the double-digit millions. The app uses tiered memberships for full access to its features, but faces competition from AI agents focused on planning and organization, along with risks tied to providing emotional support through chatbots.

Sources: TechCrunch

Read the original story ↗

Tenstorrent, a chip startup focused on inference, raised $1.5 billion in a round valuing the company at $6.2 billion. The round was led by Trace Capital and Innovation X, with participation from Cadence Ventures, Eclipse, Fidelity, SanDisk, and T. Rowe Price. TechCrunch reports that five months earlier, the company had received acquisition interest from Intel and Qualcomm at valuations of up to $5 billion.

Tenstorrent CEO Jim Keller told Bloomberg that the company decided to remain independent and raise more capital after demonstrating that its Galaxy servers could run DeepSeek models. The new capital will be used to scale chips designed to serve open-weight AI models and generate video.

Sources: TechCrunch

Read the original story ↗

MIT Technology Review argues that personal AI agents are creating a third category of web visitor: neither humans nor conventional bots, but systems that act on people’s behalf and operate at machine speed. According to the publication, Meta’s Muse reached 5 million downloads in its first month, while agentic web traffic tripled from August to September. Websites including Pizza Hut and Yelp are already presenting difficulties for these agents; Amazon has blocked Muse’s agents and is also seeking to block Perplexity’s, arguing that they threaten its advertising business. A judge recently ruled in Perplexity’s favor, but Amazon appealed.

The publication says that if demand continues to grow, platforms may prioritize agents because of their e-commerce potential. Shopify is already giving them checkout lanes that bypass pages designed for humans. The result, the analysis says, could be a less convenient web for people who still want to browse, compare products and use services directly.

Sources: MIT Technology Review

Read the original story ↗

The US Department of Homeland Security has banned DeepSeek use across civilian government agencies and ordered the technology removed from most government systems. The order also covers technology built with help from the company’s models and has been in effect since the end of July, although it was made public only on Monday. The Department of the Treasury, Department of State and Department of Justice are among the affected agencies; agencies within the Department of Defense are excluded.

The notice did not detail the security concerns. The measure comes amid continuing US-China technology tensions and targets an open-source, free and relatively inexpensive tool also used by US companies including Snowflake, Vercel and OpenRouter. DeepSeek’s popularity has grown since its public release in 2025, while Anthropic and OpenAI have claimed that Chinese AI firms, including DeepSeek, essentially copied their models.

Sources: BBC, TechCrunch, Xataka, g1

Read the original story ↗

Independent developers report that AI-generated clones are reproducing their games and discouraging them from sharing prototypes. Ars Technica points to a delayed Steam Next Fest demo as a sign of a growing problem: without showing a game, creators lose sales opportunities; by showing it, they may make copying easier. Sos Sosowski documented several complaints and described the situation as a dilemma for indie developers.

Charlie Greenman, who posted a clone of Holmér’s rotational Tetris game, told 404 Media that he saw nothing wrong with what he did. Some developers argue that greater complexity and originality can protect against superficial copies, while Holmér told 404 Media that reduced engagement with the clones has made him more comfortable taking time to develop his game properly.

Sources: Ars Technica

Read the original story ↗

Adam Mosseri, Instagram’s chief, said the platform still does not know how to combat profiles that use AI deepfakes to reproduce creators’ content, replacing their image with avatars. According to The Independent, creators have reported the problem to the executive; one case involved a comedy performance whose audio and text were kept while its image was replaced by a female avatar. Similar profiles also reportedly use dancers’ videos to generate AI characters that repeat choreographies.

Mosseri said he was “very concerned” but ruled out a broad ban on AI-generated content. He said Meta is studying ways to direct more traffic to original authors and decide how to handle complete copies. The executive also acknowledged that copyright reports do not always resolve these cases and that the platform needs to help creators without making it easier to submit false reports.

Sources: Tecnoblog

Read the original story ↗

In an interview with The Verge’s Decoder, Illinois Gov. JB Pritzker discussed artificial intelligence, technology, privacy and the role of government. Pritzker said AI could help people live to 100 but also threaten them in a way comparable to nuclear weapons; the available excerpt does not detail those possibilities.

Pritzker said some problems require collective government action and cannot be solved one person or one state at a time. He also said Illinois is negotiating with the Chicago Bears and the NFL to keep the team in the state, either in Chicago or Arlington Heights, amid a broader debate over the relationship between government and big business.

Sources: The Verge

Read the original story ↗