AI.info

The Pulse — Page 7

Browse The Pulse on AI.info.

• Nevada Data Center Fight Puts Federal Land in Governor’s Race

  A proposed data center near Boulder City has turned federal land into a campaign issue in Nevada’s governor’s race. Residents oppose the project, while Gov. Joe Lombardo and Democratic nominee Aaron Ford differ on whether data centers should be built on public land.

• Energize Lab Lists Eiliko Security Fixes; Independent Review Pending

  Energize Lab says it closed external access to its backend systems and took Eiliko’s long-term memory and custom-avatar features offline. The Verge reported that customer media remained exposed and that researcher Sammy Azdoufal said he still had SSH access.

• OpenAI Launches Astra for Law With 230 Million URLs

  OpenAI introduced Astra for Law, a GPT-6 Astra configuration built for law firms and legal technology companies. The product includes a legal search index covering more than 230 million URLs, 26 partner plugins and new controls for confidential client work.

• China’s Ad Association Issues Three Standards for GEO

  The China Advertising Association has issued three group standards covering generative-engine optimization services, results monitoring and verification, and the selection of content sources. Announced October 14, the standards are scheduled to take effect October 23.

• Deepgram Reports Up to 70% More Audio per Dollar on Trainium2

  Deepgram and AWS say a tuned version of Deepgram’s Nova-3 speech-to-text model processed 11% to 70% more audio per dollar on Trainium2. The result applies to offline batch transcription on one EC2 instance, not to every Deepgram workload.

• Zenzap Pitches AI Work Chat With $44M in Total Funding

  Zenzap is promoting an AI workplace messenger as an alternative to consumer chat apps such as WhatsApp. The company has $44 million in total funding, including a $33 million Series A that closed in August 2025.

• Aramco Closes Deal for 49% of Saudi AI Firm HUMAIN

  Saudi Aramco has completed its acquisition of a 49% stake in HUMAIN, with PIF retaining 51% ownership. The transaction includes approximately SAR 9.9 billion in cash contributions from Aramco during 2026, alongside AI, compute and digital assets.

• Cerebras Maps 3-Trillion-Parameter Kimi K3 Across 59 Wafer Chips

  Cerebras describes an illustrative deployment that maps Kimi K3 across 59 WSE-3T wafers, keeping model weights on-chip while requests move through a pipeline. The guide explains its memory and serving approach, but does not report a like-for-like power comparison for the Kimi K3 example.

• Amazon Adds New Claims to Perplexity’s Comet Lawsuit

  Amazon filed an amended complaint accusing Perplexity’s Comet AI browser of disguising automated activity and accessing Amazon customer accounts without permission. The filing arrives weeks after the Ninth Circuit rejected Amazon’s preliminary injunction under federal and California computer-access laws.

• Tempus Pays $12 Million for Recursion’s TxFM RNA Model

  Tempus will license Recursion’s TxFM RNA sequencing model for oncology-related uses while the companies extend their existing data partnership through November 2029.

• Argo-Bench Finds Data Agents Miss the Decision, Not Just the SQL

  Argo-Bench evaluates whether AI agents can investigate an enterprise-scale simulated warehouse and make decisions, not just generate SQL. Its results show that models can analyze data yet still miss the objective, while the authors caution that the benchmark simulates one city and one year.

• OpenAI Tests Sponsored Agents in ChatGPT Ads

  OpenAI is testing Sponsored Agents, a new advertising format that lets ChatGPT users start labeled conversations with businesses after clicking an ad. The company is also adding AI tools for campaign management and integrating ChatGPT Ads with HubSpot and Shopify.

• TestMu Brings Real Android and iOS Phones to Kane CLI

  TestMu AI’s Kane CLI 0.9 can run plain-English mobile tests on real Android and iOS devices in the company’s cloud from macOS, Windows or Linux. Real-device sessions require eligible plans, and the interactive live view opens in Chrome.

• Anthropic and Hg Plan European AI Consultancy, FT Reports

  Anthropic and private equity firm Hg have agreed to back an independent European AI consultancy, the Financial Times reports. The firm is expected to place engineers and consultants inside businesses, with hubs planned in London and Berlin.

• Agility’s More Than $620M SPAC Deal Clears SEC Hurdle

  Agility Robotics and Churchill Capital Corp XI say the SEC has declared effective their registration statement for a proposed merger expected to generate more than $620 million in gross proceeds. Churchill XI shareholders are scheduled to vote on the deal November 2, 2026, and Agility says the combined company would trade as AGLT if the transaction closes.

• Hugging Face Adds Open Speech and Decision Models to Apple’s Swift APIs

  Hugging Face announced AnySpeechRecognizer and AnyDecisionModel, Swift packages that let developers select speech engines and get typed, probability-based decisions on Apple devices. Its Mac tests show Parakeet beating Apple’s transcriber on two LibriSpeech sets, but the results have not been tested on iPhone and degrade sharply on overlapping meeting audio.

• Sakana AI Tests a Two-Cycle Loop for Agent Self-Improvement

  Sakana AI describes MASS, a system that uses a shared model to coordinate virtual agents, evaluate their work and train on selected runs. After two cycles, the model improved its score per output token on four research benchmarks, while results on two software benchmarks remained below the base model.

• Firefly Raises $20 Million to Unify Physical Safety Systems

  Firefly emerged from stealth on October 20 with a $20 million Series A led by Definition Capital. The startup says its platform links purpose-built hardware and AI software with existing cameras and radios to help organizations coordinate safety responses.

• Apple Sets 18+ Minimum for Dating and Relationships Apps

  Apple will require developers to declare whether apps include dating and relationship features or shared location. Apps with Dating and Relationships features will receive a minimum age rating of 18+, with a separate provision for certain location-sharing apps in Brazil.

• AI-Generated Report Nearly Sent U.S. Forces to Intercept Chinese Ship

  A U.S. military analyst used an AI chatbot to misidentify cargo aboard a Chinese ship during the war with Iran, according to four sources cited by CNN. Armed personnel prepared to board the vessel before officials discovered that the intelligence report was false.

• Pluto Security Raises $42 Million for AI Workspace Security

  Pluto Security announced on October 13 that it has raised $42 million since its 2025 founding, including a roughly $34 million latest round. The Israeli startup sells software that monitors enterprise AI tools, connected components and applications employees build with AI.

• Florida Woman Arrested After Alleged AI-Chat Threats Against Sheriff’s Office

  A Lee County arrest report says Carli Michelle Heller was arrested after investigators linked two alleged threats against the sheriff’s office to messages sent through an Anthropic AI platform. The report describes the platform’s escalation of the messages to human review and lists an October 20 court date.

• Brazil’s Election Court Seeks Answers From Meta Over AI Campaign Images

  Brazilian President Luiz Inácio Lula da Silva’s campaign filed new evidence with the Superior Electoral Court after Meta AI blocked some requests for Lula campaign images but generated similar material for Flávio Bolsonaro. The court had already ordered Meta to explain platform restrictions and preserve related records.

• Four Gemini 4 Argon Agents Beat Minecraft’s Ender Dragon

  Four Gemini 4 Argon agents defeated Minecraft’s Ender Dragon in a randomly generated Easy-mode world after 27 hours and 14 minutes. The researchers report 2,783 programs, 605 deaths and a single completed run using Mineflayer.

• MIT Finds Claude Agents Learn From Island Play

  MIT researchers let 13 Claude agents explore a simulated island for 30 hours without an assigned task, then tested what they had learned. The agents that explored generally did better on follow-up tasks, with their written memories carrying both useful techniques and mistaken beliefs.

• Johns Hopkins Notifies 3,182 Patients After Bayview AI Upload

  Johns Hopkins Bayview Medical Center is notifying 3,182 patients after a staff member uploaded a spreadsheet of patient information to an unapproved AI tool. The hospital says the file may have included treatment details and medical record numbers, and it is offering a year of credit monitoring.

• Google Releases Gemma 4 for Phones, Raspberry Pis and PCs

  Google has released Gemma 4, an open model family designed for local reasoning, agentic workflows and offline multimodal AI. The models range from edge-focused E2B and E4B versions to 26B and 31B systems for consumer GPUs and developer workstations.

• Bend uses laws and proofs to block AI coding mistakes

  Bend presents a programming language that lets developers state important rules in LAWS.bend and require proofs before AI-written code is committed. The project combines Python-style syntax, native compilation and parallel execution across CPUs and GPUs.

• Cloudflare Separates AI-Training Opt-Outs From Search

  Cloudflare introduced a Disallow AI Training setting on September 15, 2026, allowing website owners to refuse model training while keeping accountable crawlers available for search. The policy applies different controls to Search, Training, and Agent traffic and changes how Applebot, Bingbot, and Googlebot are handled.

• Ollama Adds Model Thinking Metadata to Its API

  Ollama’s v0.34.3-rc1 pre-release adds thinking controls and defaults to the metadata returned by GET /api/show, alongside three other changes.

• Odyssey Launches Odyssey-3 for Interactive World Simulation

  Odyssey released Odyssey-3 on October 8, 2026, with a research preview for developers building physical AI systems. The company says its Pro version scored 66.1 on Physics-IQ Verified’s video-to-video benchmark, while cautioning that benchmark results do not establish performance on a specific machine or task.

• Surge AI’s DAYJOB Tests Whether Agents Can Make the Right Call

  Surge AI released DAYJOB, a benchmark suite for professional agents in healthcare and finance, on September 24, 2026. The company says the top-scoring models fully passed fewer than one in four assignments in either field.

• Microsoft Research Says It Used PDD Kit to Formally Verify Login Logic in a Rust Web App

  Microsoft Research says it used PDD Kit to formally verify the login logic of Pet Store, a full-stack Rust web app. It says the kit turns plain-language requirements into formal specifications and tests, and marks which parts of the code remain trusted.

• Chatbots Miss 57% of Financial Questions in Saturn Study

  A Saturn study tested 18 AI models on 121 personal-finance questions and found incorrect answers in 57% of responses. The findings arrive as PensionBee reports that many users would act on chatbot financial advice without checking it.

• Bessent Discusses AI Incident Channel After U.S.-China Summit Deal

  Treasury Secretary Scott Bessent described a proposed U.S.-China notification process for AI incidents in an interview published October 3. The two governments had already agreed at the September summit to establish a bilateral channel and hold an AI dialogue by November.

• Atria Dawn Study Finds Humans Still Choose AI Research Methods

  A paper on Atria Dawn Preview tracks how researchers and AI agents shared work while building the model. Agents often proposed methods, but humans made most final choices about methods and project goals.

• Ayre Think Tank Study Calculates $360 Billion Difference in Potential Economic Gains Across Commonwealth Countries

  The Ayre Index estimates a $360 billion difference in potential economic gains between Commonwealth countries that meet its youth AI-readiness benchmark and those that do not. The study also points to gaps in internet use, electricity, digital capability and youth employment across member states.

• PyTorch 2.15 Makes CUDA 13.2 Its Default GPU Build

  PyTorch 2.15 ships CUDA 13.2 as its default GPU wheel, with CUDA 13.4 wheels also available and experimental Rubin support. The release removes CUDA 13.0 builds and drops binary support for Maxwell, Pascal and Volta GPUs as CUDA 12.6 builds are removed.

• Oracle Pledges $1 Billion for Partners Over Four Years

  Oracle says it will commit $1 billion over four years to expand Oracle PartnerNetwork support for companies building and delivering AI-powered services. The plan includes training, technical resources, cloud migration funding and expanded ways to sell through Oracle Marketplace, but no annual or category-level budget.

• Hochul Names Marc Gilman to New York AI Oversight Office

  New York Gov. Kathy Hochul has appointed Marc Gilman as deputy director of the state’s new AI oversight office. Gilman will help enforce the RAISE Act, including registration, safety assessments and 72-hour incident reporting for large frontier-AI developers.