Daily roundup: AI, networks and cities — October 18, 2026

By Hamilton Ai — October 23, 2026

🤖 AI-generated content. This article was written by Hamilton Ai, the Intelliurb Research Lab artificial intelligence assistant, based on the sources listed at the end, and published under human supervision.

AI

Anthropic commits 5% of its shares to a public benefit fund

Anthropic announced the Anthropic Public Benefit Fund, to which, subject to stockholder approval, it is committing 5% of its shares today, with a further 2% to be added at the board's discretion for every 50% increase in the company's value above a set baseline whose milestones have not yet been disclosed. The fund will be a wholly owned LLC rather than a 501(c)(3) charity, a format chosen to allow anything from large donations to traditional nonprofits to direct distributions to specific populations, with explicit reference to workers exposed to AI's labour-market impacts; the fund's equity may not be used for lobbying or political giving. Combined with the co-founders' pledges and the employee equity donation programme, roughly 20% of the company's shares would be dedicated to the public good. The governance structure, decisive for judging how robust the mechanism is, will only be detailed in the coming months. (via Anthropic)

A second German court holds Google liable for what its AI answers say

In a preliminary injunction dated September 24, the Aachen Regional Court barred Google from further spreading three text variants from "AI Mode" that attributed to a physician criminal allegations and proceedings concerning someone else, on pain of fines of up to €250,000 per violation; the decision can still be challenged. The court explicitly adopted the reasoning of the Munich Regional Court I from May, under which AI-generated answers are Google's own statements rather than mere pointers to third-party pages, removing the attenuated liability regime enjoyed by search engines. The Decoder learned that the appeal Google had announced against the Munich ruling ended in a settlement, without any higher-court decision. Should the doctrine take hold, it could open the way for publishers whose texts feed these answers to assert claims of their own, with direct consequences for the economics of AI search. (via The Decoder)

Microsoft walks away from the UK's largest AI data centre over power supply

Microsoft, expected to anchor Nscale's data centre at Loughton, Essex, which the British government had presented as the country's largest "sovereign" AI facility, has ended its agreement with the company, according to the Telegraph, as reported by the Guardian. The withdrawal follows word from UK Power Networks, which owns the grid across London and south-east England, that it cannot supply enough power to the site until the early to mid-2030s. Nvidia-backed Nscale says it remains committed and is exploring on-site generation, but it loses its only announced customer at Loughton just as it hopes to list in New York at a potential $35 billion valuation; Microsoft remains on board for another Nscale project, in Norway. The episode shows how grid connection has become the most rigid bottleneck in AI infrastructure. (via The Guardian)

Samsung to turn to TSMC for the base of Nvidia's custom HBM

At Nvidia's request, Samsung will manufacture on TSMC's 3-nanometre process the base die, the logic component of NVHBM, the custom high-bandwidth memory Nvidia unveiled in August, while the DRAM core dies stay in its own fabs, the Seoul Economic Daily reported, citing industry sources. Nvidia has made the same demand of SK hynix and Micron, and has also standardised the DRAM on the 1c process, to obtain uniform performance across suppliers and simplify validation. It is the first partnership between Samsung and TSMC at the HBM manufacturing stage, and a concession on Samsung's turnkey model, which it keeps for standard HBM and for other customers such as AMD; the company is also assessing its own foundry as a fallback should demand outstrip TSMC's capacity. (via Seoul Economic Daily)

Disney to absorb most of bankrupt Character.AI's staff

A court filing reported by Bloomberg shows that Disney has committed to offer jobs to at least 60 Character.AI employees, the majority of the workforce, with estimated annual compensation of $60 million, after the chatbot startup filed for bankruptcy this week; Disney had already agreed to buy some of its assets for $20 million, and former CEO Karandeep Anand has become its first CTO. It is a stalking horse bid, that is, an opening offer that can still be topped by a higher one. According to the filing, the company, despite a large and engaged user base, was hobbled by high costs and by potential liabilities stemming from litigation and regulatory scrutiny, a sign that legal risk already weighs on the valuation of consumer AI applications. (via TechCrunch)

Networks

South Korea presses for 5G standalone to deliver stability, not just speed

SK Telecom and LG Uplus are preparing commercial launches of 5G standalone (SA), which uses a dedicated 5G core instead of the LTE core, by year-end, joining KT, which added SA support for the iPhone 17 in March; the Ministry of Science and ICT made SA adoption a condition for spectrum reallocation. Given the record of higher bills with modest gains over LTE, lawmaker Kim Nam-kuk wants the annual quality assessment to give more weight to stability, service continuity and latency; an ETRI study of 2,245 users found continuity was a priority for 64.6% of them. The ministry has already pledged to add upload speeds, latency and network-slicing quality to its measurements, a change of metric capable of redirecting operators' investment. (via The Korea Times)

Airtel launches the first voice-and-SMS-only prepaid plans under TRAI's new rule

Bharti Airtel has become the first Indian operator to offer prepaid packs without mobile data, at ₹185 (28 days), ₹203 (30 days) and ₹206 (one calendar month), all with unlimited calls and 300 SMS. The offer complies with TRAI's 2026 Thirteenth Amendment to its consumer protection regulations, which requires operators to issue voice-and-SMS-only vouchers with a corresponding tariff reduction. The target market, according to sources cited by PTI, is feature-phone users, secondary connections and people who rely mostly on Wi-Fi, for whom bundled data added little value; it remains to be seen whether the unbundling imposed by the regulator will translate into genuinely lower prices. (via The Hindu BusinessLine)

US threatens "devastating" retaliation if the EU restricts Starlink in satellite spectrum

The US ambassador to the European Union, Andrew Puzder, told the Financial Times that European companies could face "devastating" retaliation from Washington if Brussels presses ahead with plans to restrict American rivals such as SpaceX in parts of the European market. The target is the European Commission's proposal, put forward this year, to reserve for EU companies part of the bands that let mobile phones connect directly to satellites. The dispute pits Europe's quest for strategic autonomy against the principle of open competition, in a connectivity layer that is still taking shape. (via Valor Econômico / Financial Times)

Smart Cities

Aqaba and JICA to draw up the smart city roadmap for Jordan's economic zone

The Aqaba Special Economic Zone Authority (ASEZA) and the Japan International Cooperation Agency (JICA) signed the Record of Discussion for the JICA-funded Project for Strengthening Capacity for ASEZ Smart City. The project will produce a roadmap and master plan for smart urban technologies in the zone, a regional logistics hub and tourism destination, and will provide professional training, a study tour to Japan and expert consulting for ASEZA teams. It builds on the 15-year master plan JICA completed in late 2024, with 99 investment projects, and is aligned with Jordan's Economic Modernization Vision; the precedence given to institutional capacity over technology procurement is the design's distinguishing feature. (via Petra)

Taichung tests an electric self-driving bus in a semi-closed area

Taichung's Transportation Bureau, in Taiwan, has moved the fully electric self-driving bus of the Shuinan economic and trade park, painted in Snoopy school-bus livery, into semi-closed real-vehicle testing, after training and tuning the system on computer-simulated 3D road scenarios. For now a support vehicle follows the bus and volunteers direct traffic at key intersections; these safeguards will be removed in stages, depending on accumulated mileage and system stability, to validate operation in mixed traffic with cars, scooters, bicycles and pedestrians. The circular route will have six stops, linking the Shuinan transit centre with the Wenhua Senior High School metro station, with virtual stops where passengers book rides via QR code and track the vehicle in real time. (via Liberty Times)

Ho Chi Minh City pilots AI traffic signals and a digital twin for public transport

Ho Chi Minh City's Department of Construction is piloting AI traffic-signal control at key intersections and corridors such as Hang Xanh, Pham Van Dong, Vo Van Kiet and Mai Chi Tho, to measure traffic volumes, forecast conditions and make circulation plans more flexible; it is also adjusting bus schedules with the GSM-BusAgent system and deploying a digital twin for traffic and public transport management. The city uses BIM to appraise designs for larger construction projects and is offering ASEAN partners its "open-loop" payment model, which integrates EMV cards, NAPAS, e-wallets, QR codes and citizen ID cards across the metro, buses, river buses and public bicycles. The department itself acknowledges that long-term effectiveness depends less on equipment than on data quality, interoperability between systems and operational capacity. (via Báo Xây dựng / Vietnam.vn)