AI.info

The Pulse — Page 9

Browse The Pulse on AI.info.

• 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.

• AWS Rebuilds AgentCore Runtime Around Elastic Memory and Snapshots

  AWS has released a new version of Amazon Bedrock AgentCore Runtime that reclaims unused session memory and restores prepared snapshots for more predictable cold starts. The company says the runtime reaches about two seconds at the 75th percentile across container images from 200 MB to 2 GB.

• US Agencies Accuse Six Chinese AI Firms of Model Distillation

  The NSA, FBI and CISA say six China-based AI companies extracted billions of tokens from U.S. frontier models through distributed, industrial-scale campaigns. The agencies urge model providers, cloud companies and API aggregators to share signals and alter responses to suspected abuse.

• Apple Previews iOS 27 and Siri AI for Developer Testing

  Apple announced iOS 27 and Siri AI at WWDC26, with developer testing beginning June 8, a public beta planned for July, and a free software update scheduled for fall 2026.

• Andromeda 2 outperforms drug-formulation baselines

  A preprint posted September 16, 2026, reports that Andromeda 2 produced more high-performing paclitaxel formulations than Andromeda 1 and a wet-lab design-of-experiments campaign. No attributable quotation appears in the arXiv abstract or metadata; the page lists authors but contains no named statement to quote.

• EPA alumni flag 30 actions that could raise AI data-center health risks

  A report from the Environmental Protection Network identifies 30 federal actions it says could increase pollution-related health risks as AI data centers expand.

• XConf Teaches Models to Calibrate From Their Past

  A new arXiv preprint describes XConf, a confidence-estimation method that uses a model’s graded past episodes to improve calibration and selective prediction.

• Databricks Pledges More Than $350 Million for Singapore

  Databricks plans to invest more than US$350 million in Singapore over three years, expanding its workforce, headquarters, training programs and startup support.

• Raindrop Announces Series A, Says It Has Raised $50M in Total Funding

  Raindrop says it has raised $50 million in total funding, led by CRV, as the startup expands from monitoring AI agents in production to testing their behavior before new code reaches users.

• FCC Exempts Certain Advanced Robots After Security Review

  The FCC says the Department of War has conditionally approved certain advanced robotic devices, removing them from the agency’s national-security Covered List. The decision creates a narrow exemption from a policy that treats foreign-produced advanced robots as an unacceptable security risk.

• Z.ai Built GLM-5.3-Flash’s Inference Stack on 100,000 AI Chips

  Z.ai says it built the production inference system for GLM-5.3-Flash across more than 100,000 Chinese-made AI accelerators. The company reports a threefold throughput improvement and says an AI coding agent helped complete the work in less than two weeks.

• Cohere Targets $3 Billion Raise at $20 Billion Valuation

  Cohere is in advanced talks to raise between $2 billion and $3 billion at a proposed $20 billion valuation, according to The Globe and Mail. The financing could include the Canadian government, existing backers and German government participation, but the terms are not final.

• Google’s ToolGrad Cuts Agent-Training Data Costs

  Google Research introduces ToolGrad, a framework that generates verified API workflows before writing user prompts. The system reaches a 99.8% pass rate while producing more complex tool-use chains with fewer tool-execution steps.

• MIT’s HardFlow Keeps AI Outputs Inside Safety Rules

  MIT researchers developed HardFlow, a method designed to help pretrained generative AI models meet strict final-output constraints without retraining.

• AIUC Raises $40M to Certify Enterprise AI Agents

  Artificial Intelligence Underwriting Company has raised a $40 million Series A led by Ribbit Capital to test and certify AI agents for enterprise use. The startup says its AIUC-1 standard evaluates agents for security, reliability, hallucinations, jailbreaks and data leaks.

• FinSkillOps Turns SEC Filing Errors Into Gated Agent Skills

  A paper on FinSkillOps presents a self-evolving multi-agent system that turns recurring SEC filing errors into scoped, tested and versioned skill patches.

• Google Opens AI Atlas as Scientists Report 7 Hours Saved

  Google has opened an interactive version of its AI & Economy ATLAS, pairing global adoption data with a new study of AI use in scientific research. The study finds that scientists save nearly seven hours a week with AI, while facing growing backlogs of hypotheses that still require physical testing and validation.

• Apple Code Shows Siri Can Delegate to Claude and ChatGPT

  Private frameworks in iOS 27 and macOS Golden Gate reveal Siri mechanisms for delegating requests to Claude or replacing Apple's server-side model with GPT-5.6. The capabilities are not yet available to ordinary users, but demonstrations show third-party models handling system actions through Siri.

• GlobalFoundries Expands SiGe Capacity for AI Optical Links

  GlobalFoundries and Marvell expand a multi-year agreement to increase silicon germanium capacity at GF’s Burlington, Vermont facility. The capacity will support optical transceivers, near-packaged optics and co-packaged optics for AI data centers.

• Gender-Coded Prompts Make AI Writing Less Formal, Johns Hopkins Finds

  Johns Hopkins researchers found that four AI systems produced shorter, less complex and less formal workplace writing when prompts used language associated with women. The study tested GPT-4, Llama, Gemma and Mistral on emails, job applications and resignation letters.

• OpenAI Reveals Habitat Storage System Serving 1 Billion Weekly Users

  OpenAI says its Habitat storage platform now serves products used by more than 1 billion people each week. The system handles more than 70 million requests per second and stores over 500 petabytes of data across almost 40 regions.

• Frontier Labs’ Researchers Challenge AI’s Race to Deploy

  A 10-day series of resignations, agent breaches and warnings from frontier AI researchers has pushed Anthropic, OpenAI and other labs to reconsider the pace of deployment. The companies now face pressure to prove that outside oversight can keep up with systems they say may soon improve themselves.

• Tempus Targets 100,000 Whole Genomes for AI Research

  Tempus AI is building a de-identified dataset of 100,000 disease-specific whole genomes linked to longitudinal clinical outcomes. The company plans to make the platform broadly available by mid-2027 and eventually expand it to one million genomes.

• OpenAI’s GPT-6 Astra Scores Face Scrutiny After Quiet Revisions

  OpenAI changed several GPT-6 Astra benchmark figures after its September 3 launch, including hallucination, cybersecurity and rival-model scores. Independent ARC Prize testing also showed a wide gap between Astra’s 99.9% result under OpenAI’s provider adapter and 62.7% under the standard harness.

• OpenAI Reports Six Model-Misalignment Incidents and Sets Disclosure Rules

  OpenAI has published six reports on unexpected or concerning model behavior alongside a framework for tracking, investigating and disclosing model misalignment. The source page contains no attributable quotation from a named speaker.

• Hugging Face Brings llama.cpp Quants to Transformers

  Hugging Face has added support for running llama.cpp quantized models through Transformers, allowing developers to load GGUF checkpoints with familiar Python and PyTorch APIs.

• Apex Intelligence Raises Nearly $50M for Self-Evolving Models

  The Beijing startup says it will use the financing to develop foundation models designed to improve their own research processes, while its claims about benchmarks and mathematical work remain unverified outside company materials.

• SpaceXAI Releases Grok Voice Transcribe 2.0 at the Same Price

  SpaceXAI says Grok Voice Transcribe 2.0 doubles the accuracy of its predecessor in real-world evaluations while keeping batch and streaming prices unchanged.

• FAA Sets September 21 AI Trial for Washington Airspace

  The FAA is preparing to deploy SMART, an AI-assisted air-traffic planning system, around Washington, D.C., under a 12-year, $875 million contract with Air Space Intelligence. The platform will analyze schedules, weather, airport capacity and airspace conditions while controllers retain responsibility for aircraft separation.

• DeepSeek V4.1 Flash Passes 1 Trillion Tokens on OpenRouter

  DeepSeek V4.1 Flash processed more than 1 trillion tokens on OpenRouter shortly after its September 10 launch. The open-weight model combines a 552-billion-parameter mixture-of-experts backbone, native vision and a one-million-token context window with sharply lower cached-input pricing.

• Treble Raises $18M to Bring Acoustic Simulation to Physical AI

  Treble Technologies has raised $18 million to expand its acoustic simulation and synthetic audio platform for physical AI applications, including robotics, wearables and autonomous vehicles.

• Anthropic Opens Biology Access to Vetted Research Teams

  Anthropic is opening applications for its Life Sciences Verification Program, which gives vetted teams access to Claude models with fewer biology-related restrictions. The beta covers drug discovery, research biology, clinical development and manufacturing, with additional review for higher-risk projects.

• LatticeFlow Finds Stronger Political Alignment in Larger Qwen Models

  LatticeFlow AI’s political-bias benchmark compares Qwen 3.7 Max with Qwen3 32B, finding the larger model closer to the Chinese pole across six China-politics categories.

• Anthropic folds Claude Cowork into chat and adds Docs, Slides

  Anthropic is merging Claude Cowork with standard chat and introducing Docs and Slides tools that work from the same conversation context.

• Anthropic Lines Up $517 Billion in Compute Deals

  Anthropic has entered agreements covering at least 14.8 gigawatts of compute capacity since October, according to an analysis by The Information. The contracts could cost as much as $517 billion over roughly a decade, although the company has not disclosed an aggregate figure and may not use all of the capacity.

• Elliptic Launches Decode AI Agent for On-Chain Investigations

  Elliptic has launched Decode, an AI agent that answers natural-language questions about crypto addresses, entities and transactions. The system shows its query, reasoning and supporting evidence, and is currently available to government customers.

• Nscale Is in Talks to Raise $3.5 Billion Before Possible U.S. IPO, Reuters Reports

  Reuters reported that Nscale is discussing a pre-IPO financing package involving convertible notes and potential Nvidia funding, while a separate Nscale-Figure compute agreement carries an initial $3.5 billion commitment.

• Bell Plans $50B-Plus AI Hub in Saskatchewan

  Bell Canada and Saskatchewan have signed a non-binding agreement to add up to 900 megawatts to Bell AI Fabric near Regina. Full development could create a 1.2-gigawatt Canadian AI hub and exceed $50 billion in total capital investment.

• Gimlet Labs Raises $300 Million for Multi-Silicon AI Inference

  Gimlet Labs has raised a $300 million Series B led by Andreessen Horowitz for its multi-silicon inference cloud. The San Francisco startup says it has billions of dollars in contracted revenue and is scaling toward hundreds of megawatts of managed capacity.