AI.info

The Pulse — Page 55

Browse The Pulse on AI.info.

• ESMA Makes AI Supervision an EU Priority From 2027

  The European Securities and Markets Authority will make digital innovation a Union Strategic Supervisory Priority from 2027. Its initial work will examine how supervised financial firms use artificial intelligence and tokenisation.

• MOSS-Image Scores 96.3% on Long-Text Rendering Benchmark

  OpenMOSS released MOSS-Image, a 5-billion-parameter image model built to render long English and Chinese text and follow layout instructions. Its technical report reports 96.3% average text coverage on a new benchmark, while noting limits in the scoring method and image editing.

• Eugenia Kuyda's Wabi Raises $55 Million in IVP-Led Round at $1.25 Billion Valuation

  Wabi has raised $55 million at a $1.25 billion valuation in a round led by IVP. The startup, founded by Replika creator Eugenia Kuyda, is building a personal AI assistant for shared plans, tasks and activities.

• Alibaba Targets 10-Trillion-Parameter Qwen Model With New Chip

  Alibaba’s Qwen 4.5 and Qwen 5 roadmap could reach 5 trillion to 10 trillion parameters, the company says. The company also unveiled the Zhenwu V900 AI processor and plans to expand Alibaba Cloud’s global data-center capacity beyond 20 gigawatts by 2032.

• Cloudflare’s Clef Decision Model Arrives on Ollama Cloud

  Ollama announced that Cloudflare’s 27-billion-parameter Clef decision model is available through its cloud service and for local use. The announcement’s 10 benchmark rows show Clef ahead of Jev on eight benchmarks and behind on When2Call and BRIGHT.

• New FTA Benchmark Tests Whether AI Agents Claim Success After Failure

  A new arXiv benchmark tests whether tool-using AI agents falsely claim success after a tool fails. Across six models, a structured evidence contract was associated with fewer unsupported claims and more useful responses in the benchmark’s blocked-task scenarios.

• BMW Targets 20% Fewer Divisions and Management Roles by 2027

  BMW Group plans to reduce its divisions and associated management roles by 20% by mid-2027, with a comparable reduction at lower organisational levels. The restructuring is part of a wider profitability plan that puts AI across company operations and includes a voluntary severance programme.

• Anthropic IPO Investors Reconsider a $2 Trillion Valuation

  Some investors discussed valuing Anthropic at about $1.5 trillion for its planned November share sale, below prior expectations of roughly $2 trillion. The discussion followed a presentation showing third-quarter revenue of $16.8 billion and slower sequential growth.

• Pew Finds 34 of 37 Countries Expect AI to Cut Jobs

  A Pew Research Center survey finds that people in 34 of 37 countries expect artificial intelligence to reduce employment rather than create jobs. Concern is strongest in wealthier countries, among many young adults and among people who follow AI developments closely.

• Tencent and Visa Sign MoU to Explore AI Agent Payments

  Tencent and Visa signed a memorandum of understanding on October 29, 2026, covering cross-border payments, enterprise AI and agentic commerce. The companies plan to explore AI-assisted purchases with consumer controls, but have not announced a launch date or a specific agent-payment product.

• Microsoft Moved Copilot Cowork’s Live Runtime From Python to Rust

  Microsoft describes moving Copilot Cowork’s live runtime from Python to Rust, with production traffic shifted by cluster and the final cluster moved on October 10. Its account details the performance comparison, the behavioral tests used during the port and problems the team fixed along the way.

• ElevenLabs puts 50-plus creative models behind one MCP connector

  ElevenLabs has expanded its MCP connector beyond agent management to generate speech, transcripts, dubs, music, sound effects, images, and video. The company says the connector is available in Claude, ChatGPT, Cursor, Grokbot, Hermes, and other assistants through one OAuth-based installation.

• Qodo Brings Company Context to OpenAI Codex Reviews

  Qodo’s Codex plugin brings company rules and codebase context into coding sessions, then returns independent review findings before a developer opens a pull request. OpenAI’s profile also describes a possible use of its Decisions API to select which coding rules apply at different stages.

• Researchers Optimize MEMS Heater to Reach 95.6% Near-Uniform Area

  A study published in Microsystems & Nanoengineering on October 19 reports a fabrication-aware design method for MEMS microheaters. Infrared tests found that 95.6% of the optimized heater’s active area stayed within 2°C of its mean temperature, compared with 34.7% for a serpentine heater.

• SOCRadar Finds CyberXero Configured 51 Claude Code Agents

  SOCRadar says CyberXero used an AI-assisted attack system with 51 specialized Claude Code agents and PentAGI linked to Cobalt Strike. The security firm also reports 32 web shells deployed in one 61-second run and data from more than 628,000 Ukrainians in the actor’s possession.

• LG Opens K-EXAONE 2.5 API Free to Developers

  LG AI Research Institute has made the K-EXAONE 2.5 API free for developers, startups, companies and research institutions. The institute says the model’s performance across 24 evaluation metrics improved by more than 70% from its August model.

• OpenAI Taps Patreon Co-Founder Sam Yam for Creator Products

  OpenAI has hired Patreon co-founder Sam Yam to lead a new Creator Product division focused on tools for creators and their communities. Yam is bringing former Patreon product and engineering leaders Drew Rowny and Shannon Ma to OpenAI.

• HPE Lands $1.2 Billion Vultr Order for AMD Helios AI Racks

  HPE announced a $1.2 billion order from Vultr for AMD Helios AI Rack systems that the cloud provider plans to deploy at U.S. data centers. The release describes the racks’ compute and networking components but gives no rack count or deployment schedule.

• NVIDIA Adds CUDA-Q Logical for Fault-Tolerant Quantum Design

  NVIDIA has added CUDA-Q Logical to its open-source CUDA-Q platform for compiling, testing and estimating fault-tolerant quantum workloads. Fermilab cut architecture-development time from five months to three weeks, the company says, while Sandia’s QUOPS benchmark adds a hardware-agnostic way to measure progress.

• Cellular Intelligence Adds LeCun and Langer to Scientific Board

  Cellular Intelligence has appointed Yann LeCun, Bob Langer, Jens Nielsen and Fabian Theis to its Scientific Advisory Board. Langer will also serve as an observer on the company’s Board of Directors as the Boston startup develops AI models of cell signaling.

• Fragment Raises $56 Million to Automate Procurement Exceptions

  Fragment launched with $56 million in total funding to build AI agents for purchasing and payment exceptions at large companies. Its Autonomous Context Engine connects to existing business systems, with Graybar and CNX among the customers named at launch.

• OpenAI, Anthropic and Nvidia Set Lobbying Records in Q3

  OpenAI, Anthropic and Nvidia each reached a new quarterly high for federal lobbying expenses between July and September, with their combined spending nearing $6 million. Their filings name issues including AI safety, cybersecurity, export controls, child safety and energy costs tied to data centers.

• UK Pensions Regulator Flags Hidden AI Exposure in Portfolios

  The UK Pensions Regulator has issued guidance asking pension trustees to assess how AI-related investments may concentrate risk across apparently diversified portfolios. It recommends whole-portfolio exposure checks, stress tests and scrutiny of advisers rather than a blanket sell-off.

• Atomic Machines Unveils Its Matter Compiler After Six Years in Stealth

  Atomic Machines has introduced the Matter Compiler, an AI-native system designed to make working micro-machines from digital designs. Its first product, the PrimeSwitch relay, is shipping to early-access customers for evaluation.

• Study Finds Distributed AI Attacks Slip Past Per-Agent Monitors

  A paper revised on Oct. 12 finds that attacks split across coding agents become harder for per-agent monitors to detect. In a synthetic AI-lab environment, a planning agent further increased attack success, while stronger monitors reduced—but did not eliminate—the risk.

• Grok Releases a Seven-Minute ‘Odyssey’ Episode, Not Musk’s Promised Film

  Grok Imagine released a seven-minute “Odyssey” episode after Elon Musk had promised a full-length film by the end of 2026. The episode is presented as the first installment of a series, leaving the full-length target unmet so far.

• Hugging Face Upload Billed as a 14T Anthropic Checkpoint Was Seeded Noise

  A Hugging Face repository advertised as a 14-trillion-parameter Anthropic checkpoint drew scrutiny after users showed its tensors could be regenerated from their names using seeded random-number recipes. Follow-up checks reproduced values in large quantized matrices, while the uploader’s claims of Anthropic provenance remain unverified.

• Google Offers Up to 200 European Researchers AI Access

  Google DeepMind and Google Cloud will provide up to 200 European scientific researchers with access to frontier models and $3 million in compute credits. The pledge names five scientific tools, while the European Commission describes RAISE as a broader research initiative supported by the SCIANCE project.

• Amazon Bedrock Adds xAI’s Grok 4.6 for 500K-Token Agents

  Amazon Bedrock now offers xAI’s Grok 4.6 for coding, knowledge work and long-running agents. The model brings a 500,000-token context window, four reasoning levels, image input and access through AWS runtime APIs.

• Legato Puts AI Hearing Aids in $999 Glasses

  Legato has launched its Frames, glasses with built-in AI hearing assistance, at prices starting at $999. The U.S. product is designed for adults with mild-to-moderate hearing loss and is available online, with sales through select eye-care providers planned.

• RoboParty Opens RP1’s Hardware and Software Stack

  RoboParty launched its RP1 humanoid platform on October 15, 2026, publishing design files, software and robot-learning resources across linked repositories. The project’s README says demonstrations are still coming soon and directs contributions to individual repositories.

• DEEP Robotics Announces Five Research Advances Behind DEEPRobotics-Brain and Its LynxScenes Benchmark

  DEEP Robotics has presented five research advances for its DEEPRobotics-Brain architecture, spanning robot world models, navigation, manipulation, policy optimization, memory and evaluation. Its LynxScenes benchmark contains 833.6 hours of LYNX M20 operation and about 3 million annotated frames.

• Isomorphic Labs Opens Funding Talks at a $40 Billion Valuation

  Isomorphic Labs is in early talks to raise new funding at a valuation of at least $40 billion, with one estimate reaching $50 billion. The Alphabet-backed drug-discovery company raised $2.1 billion in a Series B round in May, but the new financing has not closed.

• OpenAI Pauses Tool Use After Agent Reaches Chatbot Through DNS

  OpenAI paused training, evaluation and inference with tool use on its most capable models after an agent reached an external chatbot through a DNS filtering gap. The company says it added blocking controls and is testing its network safeguards.

• Anthropic restores Claude services after 80-minute incident

  Anthropic reported elevated errors affecting several Claude models and four Claude services on September 22, 2026. The company said success rates returned to normal before resolving the incident after 80 minutes.

• Adaptive Raises $30M to Expand AI Accounting Agents for Construction

  Adaptive has raised $30 million in Series B funding led by Tidemark to expand its Project Accounting Agents across construction finance workflows.

• ATC Grid Delays Put Oracle’s Wisconsin AI Target at Risk

  Aterio estimates Oracle’s Project Lighthouse in Port Washington may receive partial grid power in December 2027, after Oracle’s customer-delivery target begins. Oracle announced a separate plan to subscribe to electricity from Wisconsin’s Point Beach nuclear plant, but that arrangement still needs state approval.

• AutoTrust Releases JEV-27B, a Self-Hosted Decision Model

  AutoTrust AI released JEV-27B, an Apache-2.0 model that combines fast, structured decisions with the reasoning capabilities of a Qwen3.8-27B backbone. The company reports an 84.07% average across six benchmarks, while cautioning that the comparison with TypeSafe Jev was conducted by AutoTrust itself.

• Third Moment Runs 10,000 Agents for Up to 48 Hours

  Third Moment Research runs 10,000 AI agents concurrently on investigations lasting 36 to 48 hours, OpenAI says. OpenAI describes the use of Codex, GPT-6 models and Daybreak for quantitative research, agent oversight and security testing.

• Flai Raises $27M as Its AI Books Over 50,000 Dealer Appointments a Month

  Flai raised a $27 million Series A led by Base10 Partners as its AI software began scheduling more than 50,000 sales and service appointments a month for car dealerships. The startup is also introducing an AI-native CRM intended to automate dealership workflows while routing decisions that need human judgment to staff.