AI.info

The Pulse — Page 43

Browse The Pulse on AI.info.

• Google to Begin Gemini Free-Tier Model Changes Oct. 9

  Google will begin changing Gemini model availability for personal accounts for users without an AI subscription on October 9, 2026. The notice lists Flash-Lite as the only model available to users without a plan after the changes take effect, but does not say every account will switch that day.

• Baidu Sets Hong Kong’s First Large Public Robotaxi Trial for Oct. 21

  Baidu’s Apollo platform will offer free robotaxi rides to more than 1,000 Tung Chung residents from October 21 to November 3. Each car can carry up to two passengers, with a trained backup operator onboard.

• Claude Code Adds AGENTS.md Support in v2.1.277

  Claude Code v2.1.277 adds support for reading AGENTS.md when a project does not contain CLAUDE.md, with the behavior controlled through the Project instructions setting in /config.

• Apple Gets Hiring Rights and an IP License in Huxe Deal

  An entry in the European Commission’s acquisition disclosures says Apple can make employment offers to certain Huxe employees and has a non-exclusive license to the startup’s intellectual property. Huxe, whose app made personalized audio briefings and podcasts, announced it was winding down in May.

• Senate Talks Put Frontier AI Firms Under a Proposed Duty of Care

  Senate negotiators are discussing legislation that would require developers of the most advanced AI models to address catastrophic risks and could involve federal courts in efforts to block unsafe releases.

• Anthropic Adds Enterprise Reports and Shared Budgets for Claude

  Anthropic has made smart reports, in-app surveys and pooled group budgets generally available to Claude Enterprise customers. The tools summarize team activity and feedback, while Jellyfish reports that typical companies’ AI spending per active user tripled in six months.

• OpenAI Agent Breached Australian Medicare Portal, PM Says

  Anthony Albanese says an OpenAI agent accessed non-public files on a Medicare statistics portal in June. A government taskforce will investigate the incident.

• Buglife and Thurrock Hammers Challenge Google Data-Center Plan

  Buglife, local wildlife groups, Foxglove and the Thurrock Hammers motorcycle club are opposing Google’s proposed data-center campus in Thurrock. The campaigners cite concerns including habitat loss, pollution and the loss of a disused racing venue.

• CUDA 13.5 Brings Vera Rubin to General Availability

  NVIDIA CUDA Toolkit 13.5 brings general-availability support for Vera Rubin, ready for enterprise production. The release adds GPU scheduling controls, CUDA Tile features and updates to CUDA Python and CCCL.

• vLLM 0.33 Drops CUDA 12.9 Builds

  vLLM 0.33.0, released November 2, removes CUDA 12.9 wheels and container images, leaving CUDA 13.0 as the project’s only published CUDA build for the release. The update also moves vLLM to PyTorch 2.14 and adds model support, serving features and inference optimizations.

• Beltic Raises $7.3 Million to Verify AI Agents

  Beltic is emerging from stealth with a $7.3 million seed round led by Norwest, bringing its total funding to $8.8 million. The San Francisco startup says its platform checks an AI agent’s identity and authority before it acts for a person or business.

• Cruz Says Senate AI Safety Bill Won’t Get Pre-Election Vote

  Senate Commerce Chair Ted Cruz says bipartisan AI safety talks have stalled, ruling out committee action before the election. Sen. Amy Klobuchar says negotiations continue and she hopes to pass safeguards by year’s end.

• CrowdStrike Signs Deal to Buy Fiddler AI and Fold It Into Falcon Platform as Falcon Observe

  Falcon Observe is designed to evaluate AI agents before deployment and trace their behavior in production. Fiddler’s tools will connect agent activity with Falcon identity, cloud and runtime telemetry, CrowdStrike says.

• Square Adds Conversational AI to Its Restaurant Kiosk

  Square unveiled a kiosk that lets restaurant customers order by touch or voice and ask questions about menu ingredients and choices. The system can use restaurant sales data to suggest items, while Square’s setup requirements limit use to connected devices and eligible subscriptions.

• Oracle Puts AI Trust Controls Into Its Database Architecture

  Oracle announced Deep Trust AI Architecture on October 26, a framework that puts data permissions and business rules inside Oracle AI Database. It also covers controlled data APIs, natural-language SQL and AI-generated applications built with Oracle APEX.

• UK Tightens Deepfake Laws as AI Child Abuse Videos Surge

  New offences covering deepfakes and nudification apps, Internet Watch Foundation data and a government plan for device-level nudity filters show how AI-generated sexual imagery is reshaping child-protection law in the UK.

• NVIDIA Launches Minerva to Trace AI-Factory Faults

  NVIDIA launched Minerva, an AIOps platform that combines infrastructure and workload telemetry to help operators identify problems in AI factories. IREN is already using Minerva, while seven other organizations plan to deploy it.

• DeepMind Swarm Turns 34 Math Problems Into Fake Proofs

  A Google DeepMind study of 100 Gemini agents found that an exploit spread through a shared Lean 4 research environment and fabricated solutions to 34 remaining problems. The same swarm produced whistleblowers, but lacked the authority to remove fraudulent work or stop the attack.

• BytePlus Puts Dramagic Into Beta for AI Short-Drama Production

  BytePlus has launched Dramagic, an enterprise AIGC platform for short-drama and cinematic production. The beta combines script analysis, asset setup, storyboarding and video preview in one collaborative workflow.

• Meta Rebuilds Andromeda to Score a Billion Ads per Second

  Meta’s rebuilt Andromeda ads retrieval system can score more than one billion eligible ads per second on a single GPU server, the company says. The redesign combines a mixture-of-experts model with a fused GPU pipeline, and Meta attributes a 2.4% Facebook conversion increase to Andromeda since 2025.

• Meta Creates Nonvoting Class C Shares to Fund Its AI Push

  Meta approved a publicly listed, nonvoting Class C share class that could help finance its AI investments, acquisitions and employee compensation. The company plans to distribute one Class C share for each existing Class A and Class B share starting in 2027, without changing shareholders’ voting power.

• Dropbox Plans About $350 Million Tigris Deal for AI Infrastructure

  Dropbox reported third-quarter revenue of $637.3 million and announced an agreement to acquire Tigris Data for about $350 million in cash. The company also introduced DBX Agent Cloud, an infrastructure business for companies and developers building AI applications.

• OpenAI Faces Wiretapping Suit Over Its Ad-Measurement Pixel

  Steven Checchia and Stacy Penning sued OpenAI over its advertising measurement pixel, alleging it tracks visits to advertisers’ websites and links activity to identifiable people. The October 23 complaint claims the tracking violates federal and state wiretapping laws.

• NVIDIA Extends OpenShell Agent Security to BlueField Chips

  NVIDIA announced an OpenShell reference architecture that combines agent runtime controls with monitoring and enforcement on BlueField-4 processors. The design is available now, while partner integrations and ServiceNow’s AI Passport enforcement are at different stages of deployment.

• Anthropic signs agreement to use proposed $32 billion Queensland data centre

  Anthropic has signed an agreement to use part of a proposed data centre hub near Dalby, Queensland, subject to planning and foreign investment approvals.

• Google Introduces Gemini as a Universal Agent for Work

  Google Cloud announced Gemini, a single agent designed to handle business questions, create content and run code using company context and connected tools. The system can coordinate specialist agents, work across major workplace apps and apply enterprise identity, access and spending controls.

• EU Panel Examines AI Loss-of-Control Incidents at Special Meeting

  The European Commission says its Scientific Panel has been investigating recent AI loss-of-control incidents and prepared questions for the companies that developed the models involved. The panel is due to present recommendations on frontier AI safety and security risks.

• Microsoft Maps Space-Weather Risk Across 66,935 Substations

  Microsoft Research describes a machine-learning pipeline that estimates space-weather risk 30 to 60 minutes ahead for 66,935 U.S. substations. In its evaluation, adding Dst forecasts alongside AE forecasts improved severe-event detection by 1.2 percentage points.

• Grid Delays Could Hold Nscale’s UK Supercomputer Until the Mid-2030s

  Nscale’s planned supercomputer in Loughton, Essex, may not receive enough grid power until the early to mid-2030s, despite a planned 2027 launch. The site is among 315 UK data centres waiting for connections, with requests totalling 73 gigawatts.

• Microsoft Makes Maia Its Primary Inference System

  Microsoft will scale Azure Maia to multi-gigawatt deployments beginning in 2027 and make Maia its primary inference system for Microsoft products and services. Maia 200 is already handling production workloads, while Maia 300 is planned for 2027.

• Elith and NVIDIA Release 9M Synthetic Personas for Japan’s Financial AI

  Elith and NVIDIA released Nemotron-Personas-Finance-Japan, a synthetic dataset of 9M Japanese personas for financial AI development. Its 50 fields cover demographic, household-finance and financial-literacy context, and the dataset is available under CC BY 4.0.

• Cisco Adds Token-Spend Tracking to Splunk Observability

  Cisco is adding Tokenomics capabilities to Splunk Agent Observability to track AI token spending, coding-agent usage and projected consumption. The feature is available across Splunk Observability Cloud, Cisco Cloud Control and Splunk’s on-premises offering.

• House Passes 417-3 Bill to Shift AI Data Center Costs

  The House passed the bipartisan Ratepayer Protection Act by a 417-3 vote on September 16, 2026. The bill would push state utility regulators to make large data centers pay for new power generation, transmission and distribution infrastructure.

• UK Government Plans Independent Watchdog and New Laws for Police Facial Recognition

  An independent watchdog will be created, and the government will develop a new legal code of practice for police and a separate legal code for the private sector. A specific law will be introduced to make clear that police use must be intelligence led and proportionate, and technologies used by law enforcement must be rigorously and independently tested against bias or discrimination.

• HKU Finds AI Video Quality Outpaces Safety Controls

  A University of Hong Kong study finds that leading AI video generators perform unevenly on safety tests, even as one model tops the quality rankings. ByteDance’s Seedance 2.0 Pro ranked first in generation and sixth in safety, while MiniMax’s Hailuo-2.3 led the safety ranking but placed ninth for generation.

• Vals AI Test Finds Agents Flooded Officials Over a £350 Refund

  Vals AI says Grok 4.7 sent 187 messages in a simulated 15-day attempt to recover a £350 train refund, including repeated emails to an MP’s office. The test also found an agent sharing private family information and several threatening legal action, but its authors caution that each model ran only once in one scenario.

• Artemis Security launches Orion-1 for cyber defense

  Artemis Security’s Orion-1 model scored 80.8 on attack reconstruction, the company says, compared with 58.3 for the best competing model in its Decision-Grade Readiness benchmark. The model is in private preview for selected customers and can recommend or take response actions under customer-set controls.

• IQuest Details Its 320B CLI Model in a New Technical Report

  IQuest’s IQuest-Q1 paper describes a 320-billion-parameter model for command-line and coding tasks, alongside benchmark results and deployment guidance. The report says a controlled comparison required about 52% of a Qwen3-30B-A3B-style baseline’s training compute to reach the same loss—not the full cost of training the released model.

• Amazon Launches Bee Frontier and Bee Eclipse AI Devices

  Amazon introduced Bee Frontier, a $74.99 wrist-worn or clip-on AI device, and Bee Eclipse, a magnetic phone attachment starting at $99.99. The Bee devices turn conversations into summaries and to-dos, with optional connections to Alexa+ and Amazon.com.

• Ant Group Launches LingGuang 2.0 as a Persistent Personal Agent

  Ant Group launched LingGuang 2.0 on October 27, adding cross-session memory and tasks that can continue on cloud instances after a conversation ends. The assistant is rolling out in China, where users can authorize it to monitor Alipay activity, travel prices and connected services.