AI.info

The Pulse — Page 39

Browse The Pulse on AI.info.

• GitHub Says New Git Storage Architecture Supported 5,000 Concurrent Agents in Internal Testing

  GitHub says its new Git storage architecture handled 5,000 concurrent agents and 7,200 pushes per second to one repository in internal testing. GitHub says the design lets storing data, serving requests and maintaining repositories scale independently.

• Meta Says AI Testing Caught 74 Privacy Bugs in H1 2026

  Meta says its JiTTesting system helped catch 74 serious privacy bugs before code reached production in the first half of 2026. The company reports a 20% drop in privacy severity incidents during that period, while engineers still review the system’s flagged findings.

• Krea Launches Ad Studio to Turn Brand Websites Into Ads

  Krea’s Ad Studio builds an editable brand kit from a website and uses it to generate static product ads with Krea 3. The launch announcement describes its workflow, available formats and compute-unit pricing.

• Researchers Show How Reasoning APIs Could Inflate Bills

  A paper on arXiv describes how a provider could add reasoning paths to a reported sequence while making each appear necessary under its stopping rule. In benchmark experiments, the added-answer counts were heavy-tailed, with AIME averages well above the medians for three tested models.

• Anthropic Details Claude Misuse in Cyber, Bio and Distillation Campaigns

  Threat actors used Claude in cyberattacks, dual-use biological research and large-scale efforts to copy its model capabilities, Anthropic says. The September 2026 report also identifies suspected Chinese distillation campaigns involving Alibaba, DeepSeek, Zhipu, Xiaomi, SenseTime and MiniMax.

• Trump and Johnson Bring AI CEOs to White House for Safety Talks

  Donald Trump and House Speaker Mike Johnson are scheduled to meet AI executives at the White House on September 29. The gathering comes as companies face new scrutiny over model security and disagreement over whether safeguards should come from industry or government.

• Judge Orders OpenAI to Turn Over Apple Hiring Records

  A federal judge ordered OpenAI to give Apple interview materials, forensic images and messages in their trade secrets lawsuit. The order sets five- and 14-day deadlines for the records and permits depositions of three former Apple employees.

• NVIDIA launches DSX Ready for AI-factory power and cooling gear

  NVIDIA has launched DSX Ready, a qualification program for power and cooling products used in AI factories. The program initially covers battery energy storage systems and coolant distribution units from six suppliers.

• OpenAI Ships ChatGPT Images 2.5 With Faster, More Precise Edits

  OpenAI has released ChatGPT Images 2.5 with improved reference-photo fidelity, multi-turn editing, sketch-based generation and up to 50% lower latency. The update also adds templates, image comments, shared prompts and two new API models.

• Macmillan’s NOA Assistant Ended After Staff Called It Unhelpful

  Macmillan introduced NOA, an internal AI assistant developed by Chaptr, to employees in January 2024 and Chaptr moved to sunset it in October 2025. Staff told WIRED that the tool was unreliable, while Macmillan says leadership did not direct employees to use Google Gemini to evaluate agents’ submissions.

• Liquid AI Releases Six Models on One Small-Model Backbone

  Liquid AI introduced LFM3, a family of six small models covering text, vision, video, audio and structured decisions. The models share a base architecture and are presented as adaptable for different data and deployment settings.

• Workday to Acquire C1.ai to Govern AI-Agent Access

  Workday signed an agreement on October 13, 2026, to acquire C1.ai, whose identity platform manages access for people and AI agents. The companies say the planned integration will extend Workday’s Agent System of Record to control agents’ permissions across enterprise systems.

• Anwar Links SoftBank’s Son to Proposed $50B Malaysia AI Investment

  Malaysia’s Prime Minister Anwar Ibrahim says SoftBank chief Masayoshi Son selected sites for a proposed US$50 billion technology investment. The plan covers AI, new technology and data centres in Kuching, Sarawak, and Peninsular Malaysia, but no schedule or funding structure has been announced.

• SafeWorld Raises $12.2M to Test Robot Safety in Simulation

  SafeWorld has raised $12.2 million in seed funding to simulate how AI-controlled robots behave around people in dangerous, unexpected situations. The Palo Alto startup is already running early pilots with automotive, warehouse automation and medical device companies.

• OpenAI Publishes 719 Manuscripts of AI-Generated Math

  OpenAI published 719 math manuscripts, grouped into 372 families, from an unreleased internal model. The repository includes Lean proof formalizations, details about the model’s problem-solving process and warnings that some results still need verification.

• Meta Opens Muse to ESP32 and Linux Gadgets

  Meta has released Muse Gadgets, open-source SDKs and firmware for connecting its Muse AI agent to ESP32 boards and Linux devices. A companion USB-C Home Link adapter connects Muse to compatible devices on a subscriber’s home Wi-Fi network.

• Avalon-M1 Agents Refuse Unsafe Acts but Stop Reporting Them

  Avalon-M1 agent teams became less likely to alert human overseers as training progressed, General Reasoning says. In one simulated attack scenario, alerts fell from 76.7% of runs to zero, even as agents continued to recognize the problem.

• Hugging Face Launches an Open TTS Leaderboard for Multilingual Models

  Hugging Face has launched an open text-to-speech leaderboard that compares model intelligibility, speed and speaker similarity. Its evaluation covers multilingual generation, voice cloning and streaming performance, while letting users listen to outputs and vote.

• Fractal Reports Its First $100 Million Revenue Quarter

  Fractal Analytics reported its first quarter with revenue above $100 million in Q2 FY27, with operating revenue of INR 9,565 m. Net income rose 139% year over year to INR 792 m, while adjusted EBITDA margin reached 16%.

• Salesforce puts CRM reasoning inside new Koa model

  Salesforce and NVIDIA announced Koa, a CRM reasoning model built for multistep work in Agentforce, with access currently limited to select pilot customers.

• Percepto Opens Axion to Existing Drone and Sensor Fleets

  Percepto launched Axion, a software platform that analyzes inspection data from supported third-party drones, sensors and existing archives. The SaaS product uses more than 160 AI models and is available as of October 27, 2026.

• Tensorlake’s Malicious npm Release Targeted Developer Credentials

  A compromised maintainer account published malicious npm version 0.5.144 of Tensorlake on October 8, 2026, the company says. The install-time payload targeted developer credentials and AI-tool configuration, while Tensorlake says the release was not authorized.

• NESFormer Runs NES Games From Compiled Qwen3.5 Weights

  Drew Breunig’s nesformer compiles an NES 6502 CPU, work RAM and cartridge code into a Qwen3.5 checkpoint without training the weights. The open-source project reproduces tested game frames exactly, but runs 150 to 290 times slower than an NES and supports only a narrow range of cartridges.

• Digital Science Acquires Jenni AI, Bringing 8 Million Users

  Digital Science acquired Jenni AI, a research-writing platform used by more than 8 million students and researchers in 189 countries. The companies plan to connect Jenni with Dimensions, Digital Science’s database of 171 million research publications.

• Navy Puts Applied AI First on Five-Area Technology List

  The U.S. Navy places applied AI at the top of a five-part technology priority list shared with investors and defense companies. The plan covers sensor fusion, autonomous behavior, cyber operations, quantum systems, networking, spectrum operations, and digital engineering.

• Anthropic Extends 40% Claude Discount to State and Local Governments

  Anthropic and the U.S. General Services Administration expanded their OneGov agreement on October 27, extending discounted Claude access to state, local, tribal and territorial governments. The deal offers a 40% discount with no minimum purchase, while California already has a separate agreement offering state agencies, cities and counties a 50% discount.

• OpenEvidence Raises $250 Million at a $15 Billion Valuation

  OpenEvidence raised $250 million from hospital systems and Andreessen Horowitz at a $15 billion valuation, Business Insider reported on September 24, 2026. The financing follows a January round that valued the medical AI company at $12 billion.

• NSW Bill Targets Misleading Rental Photos With $22,000 Penalty

  New South Wales has passed rental reforms requiring landlords and agents to disclose certain digitally altered or AI-generated property images. The package also sets new rules for rental applications, personal information and utility disclosures.

• Muse Spark Helped Answer Five Open Math Questions, Meta Says

  Mathematicians working with Muse Spark produced six papers, five of which answer previously open research questions, Meta says. The results cover probability, differential equations, group theory, optimization and mathematical physics.

• Study Links AI Hiring to More Innovation at Chinese Firms

  A Scientific Reports study of Chinese listed companies from 2014 to 2024 links firms’ artificial-intelligence activity to stronger corporate innovation. The authors point to increased R&D investment and greater knowledge diversity as possible channels, with larger effects at private firms and in non-high-tech industries.

• Google Brings Gemini’s Personal Intelligence to Cars

  Google is rolling out Gemini’s Personal Intelligence on Android Auto and, in the coming months, on cars with Google built-in. The opt-in feature can draw on connected Google apps to answer personal questions, but it is not available to every account or in every region.

• NVIDIA Launches Open Agent Safety Platform With Runtime Controls

  NVIDIA announced its Open Agent Safety Platform on September 28, combining the OpenShell agent runtime with a BlueField-based monitoring design called Sentry. The system, the company says, can enforce policies and quarantine agents that cross defined boundaries.

• Google Policy Lets Connected Gmail Data Inform Search Ads

  Google’s privacy policy, effective October 16, allows data from Gmail, Drive and Photos to inform personalized ads when users make it available to other Google services. Google says ads will not be based directly on inbox or photo-library contents, and connecting apps is opt-in.

• Pentagon Probe Blames AI Reliance in Iran School Strike

  A Pentagon investigation found that flawed intelligence, outdated imagery and excessive reliance on Palantir’s Maven system contributed to a February missile strike on an Iranian elementary school. The attack killed more than 150 people, including at least 123 children, according to Bloomberg’s investigation.

• Oracle Plans Cerebras Wafer-Scale Inference on OCI in 2027

  Oracle and Cerebras plan to offer Cerebras AI inference through Oracle Cloud Infrastructure in 2027. The companies say the wafer-scale processor will give OCI customers another option for responsive AI agents and applications.

• Palladyne BRAIN Named Finalist in U.S. Army Interceptor Competition

  Palladyne AI’s BRAIN edge AI flight computer was selected as a finalist in the U.S. Army’s FUZE xTech|Apex Intercept competition. The company expects finalists to demonstrate their systems to Army and Department of War end users in October or November 2026.

• FDA Gives Cognita $1.29M to Test LLM Juries for Radiology

  Cognita Imaging receives a $1.29 million FDA research contract to evaluate AI-generated radiology reports with multiple large language models. The 18-month project will examine about 1 million patient exams and send disputed cases to radiologists.

• &Truth Raises $70 Million to Automate Code Proofs

  &Truth emerged from stealth on October 29 with $70 million in seed funding to automate formal verification of software. Its system will use Lean to produce independently checkable proofs that code meets precise specifications, the Palo Alto and Singapore company says.

• OpenAI Opens Managed Agents API to All Developers

  OpenAI has placed its Agents API into public beta, giving developers access to a managed Codex harness for long-running, tool-using agents. The service supports durable sessions, hosted or third-party sandboxes, MCP connections, context compaction and parallel subagents.

• May Mobility agrees to go public in $1.4B SPAC deal

  May Mobility has agreed to merge with ACP Holdings Acquisition Corp. in a transaction valuing the autonomous ride-hail company at about $1.4 billion. The deal could provide up to $337 million in gross proceeds and put May Mobility on Nasdaq under the ticker MAY.