AI.info

The Pulse — Page 10

Browse The Pulse on AI.info.

• Amazon Weighs a Reported $7 Billion Decart Acquisition

  The Wall Street Journal reports that Amazon is in talks to acquire AI startup Decart in a deal that could be valued at around $7 billion. Decart’s own materials describe a business spanning software for AI-chip efficiency, live video and world-model products, while identifying Amazon as an existing strategic customer.

• Google Study Finds AI Raises Patent Quality but Splits Skill Gains

  A three-month trial of 133 patent lawyers found that Google's AI drafting assistant improved work quality, but gains in unaided judgment appeared only among senior lawyers. Junior lawyers showed no average skill improvement after the tool was removed, and their scores spread toward both stronger and weaker results.

• NTT DATA agrees to deploy 900 MW of Bloom Energy power at data centers

  NTT DATA has agreed to deploy 900 megawatts of Bloom Energy fuel-cell power across data center projects worldwide, with initial shipments starting in 2027. The companies’ relationship is designed to expand across NTT DATA’s global data center portfolio and reach gigawatt scale.

• San Francisco and Oakland Approve 45-Day Data Center Pauses

  San Francisco and Oakland approved 45-day pauses on new data center approvals on October 6, 2026. The cities say the temporary measures will give officials time to examine energy, water, environmental and community impacts.

• REA Connects Coding Agents to Reverse-Engineering Tools

  REA connects coding agents to tools for inspecting binaries, apps and websites through MCP and a command-line interface. Its latest project update describes agent-assisted game reconstructions, while the software returns evidence and limitations alongside its findings.

• Marine Corps Picks SteerBridge for Aircraft AI Prototype

  The Marine Corps selected SteerBridge Strategies to help prototype AI and machine-learning tools for aviation supply, predictive maintenance and planning. The selection followed a 34-day process but is not a contract award; work depends on completing a statement of work and other acquisition steps.

• McDonald’s Faces Antitrust Suit Over AI Pricing Recommendations

  A proposed federal class action accuses McDonald’s of using an AI-enhanced pricing tool to share sensitive sales data among franchisees and weaken price competition. McDonald’s says franchisees set their own prices and its optional tools do not coordinate them.

• Cantwell Wants Independent Audits Before Frontier AI Releases

  Sen. Maria Cantwell outlined a six-part federal framework for frontier AI on October 7, 2026. It calls for NIST standards, independent audits before covered models are released, incident reporting and protections for people affected by AI decisions.

• AWS Publishes Production Blueprints for NVIDIA Cosmos 3

  AWS has published reference architectures for running NVIDIA Cosmos 3 across data curation, fine-tuning, cloud inference and edge deployment. The plans combine Amazon S3, SageMaker HyperPod, EKS, JumpStart and IoT Greengrass, with different GPU recommendations for each workload.

• Ollama Adds DFlash and Image Input to Its Apple MLX Engine

  Ollama’s August 10 announcement adds DFlash acceleration and image input for Muse Glimmer on Apple Silicon.

• Tekever Raises $580M at a $6.4B Valuation

  Tekever has closed the first part of a $580 million Series D round at a $6.4 billion valuation. UC Investments and Baillie Gifford led the financing as the autonomous-systems company prepares to expand production, pursue acquisitions and deliver a major British Army surveillance program.

• CERT-In Says AI Agents in Testing Reached Indian Live Systems

  India’s CERT-In says AI agents used in research and evaluation reached live systems at Indian organizations and read business records containing personal and confidential information. A separate financially motivated campaign used open-source agentic tools to exploit familiar security weaknesses faster than human attackers.

• Deepslate Raises €7.7 Million to Build European Speech AI

  Berlin-based Deepslate has raised €7.7 million in seed funding led by 42CAP to develop its speech-to-speech AI models. The company plans to expand European-language training data, improve its models and scale infrastructure in European data centres.

• Epic Delays Disney Shooter as It Shifts to AI-First Development

  Epic Games has delayed its Disney collaboration, an unannounced shooter featuring Marvel and Star Wars characters, and directed its team toward AI-first production. Bloomberg reports the project was planned for March 2027 and now has no announced release date.

• Florida Sues A2E Operator Over AI-Generated Images of Minors

  Florida Attorney General James Uthmeier sued Rengstorff Consulting, alleging its A2E image platform lets users create sexualized images of children and non-consenting adults. The state alleges the company’s safeguards failed to stop misuse and is also investigating possible criminal offenses.

• SB Intuitions Adds Reasoning and Tool Use to Sarashina

  SB Intuitions introduced Sarashina3.2 mini thinking on October 29, adding adjustable reasoning, web search and multi-step tool use to its Japanese-developed model series. The company reports strong agent-benchmark results but also acknowledges weaker scores in Japanese culture questions and gaps against larger rivals.

• Anthropic Finds Robots Can Do 74% of Physical Tasks, but Cost Too Much

  Anthropic researchers estimate that robots can perform 74% of physical tasks in the United States, but are cost-competitive for just 0.3% of job tasks. The study says robot-cost estimates often include human supervision, exception handling and repair work, while identifying other barriers to adoption.

• Terminal-Bench-Science Puts Research Agents Through 70 Tasks

  Terminal-Bench-Science 0.1 selects 70 research workflows from 920 proposals, with tasks graded by their own tests. An independent Artificial Analysis evaluation reports separate results using its own harness.

• Liquid AI Releases Open d1 Models With 16ms Jetson Decisions

  Liquid AI released d1-3B and experimental d1-omni-600M as open-weight multimodal decision models on October 7, 2026. The company reports d1-3B answers a single question in 16 milliseconds on an NVIDIA Jetson AGX Thor, while noting that the smaller model has no published speed results.

• Geoffrey Hinton Gives Congress One Year to Rein In AI

  The Nobel laureate warned lawmakers that they may have only about a year to establish safeguards before artificial intelligence becomes harder for humans to control.

• NVIDIA's Reshard Refit Achieves a 2.7x to 5.7x Speedup Across Various NeMo RL Benchmarks

  NVIDIA says its Reshard approach transfers only the theoretical minimum amount of model-weight data between training and inference GPUs. In NeMo RL benchmarks, Reshard refit achieved a 2.7x to 5.7x speedup over conventional gather-and-broadcast refit.

• Sony Brings QSSR AI Upscaling to the Standard PS5

  Sony announced Quick Spectral Super Resolution (QSSR) for the standard PS5 on October 1, describing it as a new performance tier of AI upscaling. Marvel’s Wolverine and Ghost of Yōtei are the first games to receive QSSR as an optional graphics setting.

• Lake Mariner Fire Exposes Accountability Gaps in $3.2B AI Campus

  A June fire at TeraWulf’s Lake Mariner data center in New York exposed gaps in emergency planning, water access and corporate accountability. The $3.2 billion project involves TeraWulf, Fluidstack, Google and Anthropic, while state officials weigh tougher data-center rules.

• Elorian Launches Glimpse to Reason Through Images and Video

  Elorian has released Glimpse as a research preview, describing a model that can revisit visual evidence while reasoning about images and video. The company reports results across multimodal evaluations and says direct action outputs for robots remain a work in progress.

• White Circle Finds DALL·E 3 Guardrails Fail in 13% of Attacks

  White Circle’s RISE method produced human-confirmed explicit images in 13.33% of tested attack prompts against DALL·E 3, with lower rates on Google’s Nano Banana 2 and OpenAI’s GPT-Image-2. The researchers say a newer version of the method also found weaker safeguards in recently released image models, though they published no rates for those tests.

• Codex Security Finds Node.js Brotli Flaw; Imperva Confirms RCE

  OpenAI Codex Security researchers Khai Tran and Sylvie Mayer reported a high-severity flaw in Node.js’s bundled Brotli decoder. Imperva later demonstrated code execution on a vulnerable official Node.js build and confirmed the patched release rejected the same input.

• DayOne Adds RM15 Billion for Johor Data Center Buildout

  DayOne has arranged RM15 billion in new financing from seven banks for its data center platform in Johor, Malaysia. The package includes a US$2.4 billion term loan and a RM5.5 billion Murabahah facility, bringing its financing platform to RM30 billion.

• Cognizant’s 75M-Parameter Model Hits 76% on ARC-AGI 1

  Cognizant AI Lab reports that a 75-million-parameter transformer reached approximately 76% pass@2 on ARC-AGI 1’s public evaluation set after a little over four hours of training and evaluation on one NVIDIA H100 GPU.

• Reference-Tenant Notice Sets Mid-November Target for Teams Deepfake Alerts

  A 365Fanatic reference copy of a Microsoft Message Center notice says targeted rollout of third-party synthetic audio and video detection in Teams is expected to begin in mid-November 2026. Microsoft’s Graph documentation describes a separate beta API for reporting detections and does not announce the rollout schedule.

• IQVIA Launches Models That Simulate Clinical Trials

  IQVIA introduced Life Science Models at TechIQ 2026, including tools to simulate studies and adjust trial protocols. The company says 85% of predictions in an initial evaluation across 50 trials and 270 endpoints were on target or clinically close.

• Microsoft exits Nscale’s Essex AI data centre deal

  Microsoft has withdrawn from its planned role as anchor customer at Nscale’s Loughton AI data centre in Essex, citing power supply concerns. Nscale says it remains committed to the site and is exploring alternative power sources after a grid connection delay pushed expected supply into the early to mid-2030s.

• U.S. Courts Weigh Burden Shift for Deepfake Evidence

  Federal evidence officials are considering a new authentication standard for evidence challenged as AI-generated or altered. The proposal would require an initial showing from the opponent before shifting the burden to the evidence’s proponent.

• Stanford Deploys 37,000 AI Agents to Design Drug Candidates

  Stanford researchers created a virtual biotech company staffed by thousands of AI agents that analyzed clinical trials, identified biological features associated with drug-development success and proposed a lung-cancer therapy later independently developed by a pharmaceutical company.

• King Charles Brings Four AI Firms Into Scottish Safety Talks

  King Charles III is meeting representatives from OpenAI, Anthropic, Google DeepMind and Nvidia at Dumfries House in Scotland. The talks focus on keeping advanced AI aligned with human welfare, community interests and the natural world.

• Ubitium Raises $14 Million for a Universal Physical-AI Chip

  German chip startup Ubitium has raised $14 million in seed funding led by Ibex Investors to develop a processor for physical AI. Its first chip demonstrated signal processing, AI inference and real-time control on one reconfigurable array; evaluation units for its planned commercial processor are expected in 2027.

• Meta Says Muse Was Heavily Inspired by OpenClaw

  Meta product chief Nat Friedman says the company built its Muse AI agent from scratch but took heavy product inspiration from OpenClaw. The admission follows user discoveries of matching workspace filenames and near-identical SOUL.md configuration content.

• Microsoft Reports $10 Billion in Revenue From OpenAI Arrangements

  Microsoft recorded $10 billion in revenue from commercial arrangements with OpenAI during the quarter ended September 30, 2026, and reported $7.2 billion in accounts receivable from OpenAI at quarter-end. The company also reported $63.9 billion in Microsoft Cloud revenue and 48% Azure revenue growth for the quarter.

• AP-NORC Poll Finds 64% Say AI Is Moving Too Fast

  An AP-NORC poll found that 64% of U.S. adults think AI is developing too quickly, including 76% of Democrats and 54% of Republicans. The survey also found broad support for government action to keep AI under human control and protect workers.

• Zoom Adds AI Tools for Sales Outreach and Forecasting

  Zoom introduced Engage and Forecast as new capabilities for its Revenue Accelerator platform on September 15, 2026. The tools combine buyer signals, customer conversations, sales outreach, and deal forecasting under Zoom’s proposed AI-powered revenue operating system.

• OpenAI adds three state-policy leaders as AI rules multiply

  OpenAI is adding Jessica Schumer, Caulder Harvill-Childs and Thomas MacLellan to its global affairs team as states advance new AI laws. The hires support the company’s strategy of shaping state rules while Congress struggles to establish a national framework.