AI.info

The Pulse — Page 25

Browse The Pulse on AI.info.

• Musk Drops Apple Claims and Keeps Antitrust Case Against OpenAI

  X Corp. and SpaceXAI ask a Texas federal court to dismiss their antitrust claims against Apple after resolving the dispute. The companies keep their claims against OpenAI over Apple’s integration of ChatGPT in iPhone features.

• HSBC Plans Hong Kong AI Institute as CEO Warns of Job Losses

  HSBC plans to open an AI and emerging technology institute in Hong Kong in early 2027, with research focused on areas including AI safety, staff training and fraud detection. At Hong Kong FinTech Week, chief executive Georges Elhedery said AI would transform jobs, eliminating some while creating others.

• IBM Puts GPT-4.1 First on τ²-Bench Retail With Learned Memory

  IBM Research’s ALTK-Evolve system lifted a GPT-4.1 customer-service agent to a 67.5% Pass⁴ score on τ²-bench retail, topping the leaderboard snapshot it compared against, the team says. The system adds guidelines learned from past agent runs to the prompt, without changing model weights, but raises reported retail inference costs.

• Google Cloud Sets November Launch for Vera Rubin A5X Instances

  Google Cloud and NVIDIA announced A5X cloud instances built on Vera Rubin NVL72 systems, with general availability planned for November 2026. The companies say the infrastructure will support AI agents and reinforcement learning, with Google targeting clusters of up to 960,000 GPUs across multiple sites.

• OpenAI Gives Ukraine Daybreak Access for Civilian Cyber Defense

  OpenAI is extending its Daybreak cybersecurity program to Ukraine’s government to help defend civilian infrastructure. The initiative will support vulnerability discovery, software analysis and testing of security fixes amid nearly 6,000 cyber incidents handled by CERT-UA in 2025.

• Amazon Lets a Subset of Customers Text Alexa for Shopping

  Amazon is inviting a small group of U.S. customers to text Alexa for Shopping, with access expected to expand to millions more in the coming weeks. The service can help with grocery orders, product searches, price alerts and purchases, and Amazon says requests can carry over between text, its website, its app and Alexa-enabled devices.

• Helsing Unveils Contra and Two New Air-Defense Interceptors

  Helsing has unveiled Contra, an AI-based air-defense system, alongside its AD-1 Inferno and AD-2 Ambush interceptors. The weapons remain in development, with availability no earlier than the second half of 2027.

• White House AI Task Force Starts Its 120-Day Review

  The White House task force adopted its charter on October 15 and began a 120-day review of federal AI policy. Its work includes risk assessments, breach-reporting protocols and coordination on the White House Accord on Super Intelligence.

• Meta bundles paid AI usage into $7.99 Meta One plans

  Meta has launched Meta One, a subscription service bundling extra features across Instagram, Facebook, WhatsApp and Meta AI. Individual plans start at $7.99 per month and include expanded access to image, video and in-app AI tools.

• Manus Seeks $500 Million at $4 Billion Valuation After Meta Split

  AI startup Manus is discussing a $500 million funding round that would value the company at about $4 billion. The Singapore-based company is also weighing a Hong Kong listing after resuming independent operations following the collapse of its Meta acquisition.

• Sierra Launches Fleming-1 to Flag AI Callers in Real Time

  Sierra says fleming-1 analyzes phone-call speech in real time to identify callers likely to be AI agents. The model flags suspected AI calls, leaving businesses to decide how to respond.

• Vera CPU Runs 1.6x More Agents per Watt Than the Fastest x86 CPU, NVIDIA Says

  NVIDIA’s Vera CPU supports 1.6 times more agents per watt than the fastest x86 CPU, the company says, in a comparison based on a 100-megawatt data center. Results from six early adopters, including tests of agentic coding, orchestration, ClickBench, TPC-H and agent sandboxes, are also reported.

• SecondState Releases FAB, a 50-Task Test for Finance Agents

  SecondState’s Finance Agents Benchmark gives AI agents 50 due-diligence tasks on a synthetic distributor’s 160-file data room. The best of four models met every criterion on 60% of answers; none passed more than 24 tasks in all three trials.

• Nous Research raises $90 million at a $1.5 billion valuation

  Nous Research, the maker of the open-source Hermes Agent, said it raised $90 million in a Series B to build a business version of Hermes. The Wall Street Journal and TechCrunch put the valuation at $1.5 billion, with Robot Ventures leading. In July, TechCrunch had reported the company was seeking at least $75 million.

• Huawei Cloud Expands Agentic AI Infrastructure Beyond China

  Huawei Cloud’s Agentic Infra platform now serves more than 3,500 customers as the company launches AI Cluster Service globally. The company also introduced new memory, model access, agent development, and industry tools at HUAWEI CONNECT 2026.

• Black Forest Labs Launches FLUX 3 Image With Box-Level Editing

  Black Forest Labs launched FLUX 3 Image with targeted image editing, layouts controlled by bounding boxes and outputs up to 4K. The company is offering 50% off API use through October 8 and says an open-weights version will follow in the coming weeks.

• Intel and Oracle Plan Xeon Expansion for OCI Agentic AI

  Intel and Oracle plan a multiyear expansion of Xeon-based computing on Oracle Cloud Infrastructure for enterprise applications, AI agents and inference. Planned OCI X14 offerings would use Intel’s next-generation Diamond Rapids processor, but the companies have not committed to a delivery date.

• Meta Restores Muse Sharing After Replacing Artifact Links

  Meta has restored sharing in its Muse AI assistant after Claroc researchers found more than 6,800 share hostnames through passive DNS. Previously shared artifacts now have new URLs, while old links direct visitors back to the person who shared them.

• Evervault Launches Browser Tokens for AI Agent Checkouts

  Evervault introduced Browser Tokens at Money20/20 in Las Vegas on October 20, 2026. The product lets browser agents use payment cards and credentials without passing their plaintext values through the agent or browser platform.

• One in Four Math Preprints Disclosed AI Use in August

  Epoch AI found that 25.25% of math preprints submitted to arXiv in August 2026 acknowledged some form of AI use. The analysis also found that 5.86% credited AI with a substantial research contribution, while warning that the figures measure disclosures rather than all AI use.

• Wikimedia Finds Unauthorized Activity It Links to OpenAI Agents

  The Wikimedia Foundation says agents it believes OpenAI operated made unapproved wiki edits, probed its Etherpad service and generated heavy traffic. The activity may have contributed to a partial Wikidata Query Service outage in May, but Wikimedia found no evidence of compromised systems or data.

• Modal Makes Multi-Node GPU Clusters Generally Available

  Modal opened its multi-node GPU clusters to all workspaces on October 1, 2026, with gang scheduling, per-second billing and RDMA networking. The service uses a Python decorator to request coordinated GPU nodes, but requires full-GPU nodes and remains bounded by each workspace’s plan limits.

• France Summons Platforms as AI Fakes Spread During Protests

  French officials met social-media representatives after protests and unrest, while a separate meeting secured marketplace commitments to help curb the resale of goods suspected of being stolen. The online coverage included both widely shared footage of violence and AI-generated images, including a Pink Panther-costumed figure.

• Agility unveils Digit 5 for barrier-free industrial work

  Agility Robotics has unveiled Digit 5, a humanoid robot designed to work near people without traditional physical safety barriers. The robot can lift 50-pound loads, work more than 20 hours per day and begin early customer access in 2027, the company says.

• Nvidia Sets a 51% Non-China Sales Threshold for Cloud Clients

  Nvidia has told cloud computing partners to direct at least 51% of their services to customers outside China, the Financial Times reports. The reported threshold applies to future contracts, while an earlier account described Nvidia’s guidance as nonmandatory.

• Anthropic Asks D.C. Circuit to Rehear Pentagon Risk Ruling

  Anthropic asked the full D.C. Circuit on October 13 to reconsider a 2-1 ruling upholding the Pentagon’s supply-chain risk designation. The company argues the decision misreads federal procurement law and allows retaliation for its AI-safety positions.

• Florida Seeks to Halt OpenAI Model Development Without Safety Approval

  Florida Attorney General James Uthmeier asked a court on September 28 for a temporary injunction restricting OpenAI model development without independent safety guardrails. The motion also seeks limits on minors’ access to ChatGPT and on how the company describes and designs the service.

• Semperis Survey: AI Featured in 84% of Ransomware Attacks

  In a Semperis survey, 84% of ransomware-hit organizations experienced attacks involving AI, including attacks that manipulated victims’ own AI systems. Its 2026 report also finds repeated attacks, threats against employees and families, and gaps in identity-system recovery planning.

• OpenAI’s More Than 100 Math Results Reopen a Credit Fight

  OpenAI says a new internal model has resolved more than 100 long-standing math problems, as mathematicians question how the company releases results and recognizes prior work. The debate follows an August meeting and a separate dispute over OpenAI’s Navier–Stokes announcement.

• Meta details TLX, a Triton extension for GPU orchestration

  Meta has described TLX, an extension to Triton that gives developers more direct control over how GPU work is organized. It has used TLX in production workloads including recommendation-system training and inference, and large-scale LLM pretraining.

• fal Launches FLUX 3 Max for Faster AI Video Iteration

  fal’s FLUX 3 Max model can generate a 10-second video in 10.95 seconds, the company says, while scoring 1,139 Elo in preliminary human-preference evaluations. The model, developed with Black Forest Labs, is available through fal’s API for text-to-video, image-to-video and keyframes-to-video generation.

• TD’s Credit Assist Cuts Commercial Memo Drafting to 30 Minutes, Bank Says

  TD Bank U.S. launched Credit Assist, a multi-agent AI system that drafts six sections of commercial credit memos in less than 30 minutes. The bank says it tested the tool on 220 curated deals, with 95% of evaluated outputs passing expert review.

• Tolan Raises $30 Million to Expand Its AI Companion

  Tolan has raised $30 million from Union Square Ventures, Khosla Ventures and Homebrew. The AI companion app has about 250,000 monthly users, the company says, and plans to introduce a free tier.

• CME Revises Compute Futures Effective Date to ‘To Be Announced’

  CME Group has changed the effective date for its planned H100 and B200 compute futures listings to “to be announced,” after first targeting October 5, 2026. The proposed contracts would track Silicon Data GPU rental indexes, while a CFTC request for comment addresses oversight and other issues in compute derivatives markets.

• Claude Code 2.1.278 Makes Server-Side Auto-Mode Classification the Default

  Claude Code v2.1.278 changes auto mode for Claude API, Enterprise, Bedrock, Vertex, Foundry and gateway deployments, adds a status indicator and warns on billed fallback.

• OpenAI Offers Academic Labs 30 Free Seats for GPT-Rosalind

  OpenAI is offering eligible U.S. university labs up to 30 free ChatGPT Business seats for 12 months through its new Rosalind Academic Plan. GPT-Rosalind access still requires a separate trusted-access approval, and the plan does not include API access.

• Mistral Unveils Large 4, a 1.05 Trillion-Parameter Model

  Mistral has opened a public API preview of Large 4, a 1.05-trillion-parameter model, while planning to release its weights on October 27. Mistral will publish further details on the model before the weights become available.

• DeepSeek Nears at Least 80 Billion Yuan in Funding Ahead of IPO

  DeepSeek is close to securing at least 80 billion yuan in new funding, exceeding its initial target of about 50 billion yuan as it prepares for a planned early 2027 IPO. Bloomberg reports that Tencent and CATL are among the largest backers, while the final amount could approach 100 billion yuan.

• NIK C2 Malware Routes OpenAI Prompts Through Infected PCs

  NIK C2 is a Windows command-and-control framework that sends an operator’s prompts to OpenAI from an infected computer, Cisco Talos says. If the operator supplies no API key, the malware can use a key stored in the victim’s OPENAI_API_KEY environment variable.

• DoorDash Lets U.S. Customers Order Food by Text

  DoorDash is testing a text-based ordering assistant that can find restaurants, build a cart and handle checkout inside a text conversation. U.S. customers using iOS can join a waitlist for the beta, which builds on the company’s Ask DoorDash assistant.