Academic libraries across North America, Europe and Oceania report being hit by waves of data-scraping bots seeking content for AI companies' large language models. Staff at institutions including San Jose State University and University of Victoria say the bot traffic has slowed or crashed digital archives, disrupting access for students, researchers and the public. Researchers say the problem emerged in late 2024 and has been worsening since.
cbc.ca
· 2026-11-06
Startup Underdog launched Woof, a 31.7MB text-to-speech model with 31 million parameters that runs entirely in a browser without internet access. The company showcased it through a retro-themed browser game where users can type text for the model to speak and manipulate a music tuner, all voiced by Woof. Underdog made the model open for download, and the demo follows the company's earlier release of Bark, a 167MB speech-to-text model for edge devices.
techcrunch.com
· 2026-11-06
Google Threat Intelligence Group and Mandiant report that cybercriminals are stealing access to commercial large language models through compromised cloud credentials and reselling that access on underground markets. In some cases, fraudulent usage charges topped $100,000 in a single day, with total costs to a victim organization reaching $1 million before the compromised access was cut off.
darkreading.com
· 2026-11-03
A proof-of-concept project encodes US federal individual income tax forms and instructions as YAML files, each line tied to a citation of the archived IRS source page it was drawn from. A build process and a citations checker verify that every line, parameter and gate has a traceable citation, and the software uses this data to run a self-preparation interview that computes values without ever filing or transmitting return data. Much of the encoding was produced by language-model agents, but the authors say the structure lets anyone verify its citations independent of who wrote it.
github.com
· 2026-10-28
A research team shared updates on H-Nets, an architecture introduced in 2025 designed to learn language directly from raw bytes rather than relying on hand-crafted tokenizers. The post is the first in a two-part series describing new results and improvements made to the technology since its initial release.
cartesia.ai
· 2026-10-28
Artificial Analysis has released version 5 of its Intelligence Index, the widely cited independent metric ranking frontier AI models, following criticism of flaws in prior benchmark aggregation methods. The piece discusses how AA's index and competitors like Epoch AI's Capabilities Index attempt to distill model performance into a single comparable score, despite inherent trade-offs in how such scores are aggregated.
interconnects.ai
· 2026-10-28
A journalist used large language models to 'vibe-research' a fabricated scientific paper, generating hypotheses, code and text with minimal human input, then worked with Jenny Mahoney, managing editor of IEEE Access, to test whether the fake submission could slip past the journal's review defenses. The experiment follows data showing a sharp rise in detectable AI involvement in academic papers, from about 12 percent in 2023 to an estimated 57 percent in 2025, alongside reports that over 12 percent of Nature Communications reviews in 2025 were entirely AI-generated.
spectrum.ieee.org
· 2026-10-27
Bitcoin Core merged a policy change on October 22, 2026, allowing fixes for vulnerabilities that are easy to rediscover to be backported and disclosed within days rather than after a year-long embargo. The change, authored by developer Antoine Poinsot, explicitly cites AI tools' ability to resurface old bugs, and follows research showing 43.6% of sampled network peers run unpatched software.
techreport.com
· 2026-10-26
Cyber insurance providers such as Coalition, Chubb, AXA XL, Cowbell and Resilience are broadening their role beyond writing policies, taking a more active hand in clients' cybersecurity defenses. The shift comes as AI tools let attackers craft faster, more convincing phishing and social engineering campaigns, straining insurers' historical loss models built around traditional ransomware patterns. Cowbell's Thomas Pytel told Dark Reading his firm now favors underwriting that continuously adapts to changing risk.
darkreading.com
· 2026-10-23
Flapping Airplanes, a one-year-old AI lab, announced a $1 billion Series A round co-led by Index Ventures and Lux Capital at a reported $6 billion valuation. The raise follows a $180 million seed round nine months earlier that had valued the company at $1.5 billion, with Sequoia, GV and Kleiner Perkins returning as investors alongside new backers including Nvidia.
techcrunch.com
· 2026-10-21
Nonprofit Earth Species Project applied machine learning to detect carrion crow vocalizations and identify which bird made which call. The group is now developing NatureLM, a model trained on human speech, music and animal sounds, aiming to interpret vocalizations across species, building on decades of recording tools like hydrophones and biologgers.
technologyreview.com
· 2026-10-21
Researchers led by Heydari et al. describe in Nature an AI system that analyzes an individual's personal health data to generate tailored health advice, and propose methods for evaluating such tools. The work responds to surveys suggesting roughly one-third of US adults and about 230 million people worldwide weekly consult chatbots like ChatGPT for health questions.
nature.com
· 2026-10-21
GoKawiil