AI News
AI news for practical tool users.
Track useful AI product updates and workflow trends without chasing hype.
Tempus expands BMS data and AI research access, with Pathos adding neurology work
Tempus and BMS expanded oncology data and Lens access through 2031, while separate Pathos agreements cover neurology records and agentic analysis.
Pony.ai, inDrive and NVIDIA plan Kazakhstan robotaxi feasibility project
An October 23 memorandum proposes Gen-7 robotaxis for Astana, with government agreement and operational assessment still required before service begins.
Sakana unveils Aquarium to connect enterprise rules with AI agent workflows
Sakana AI’s Aquarium organizes business knowledge, permissions and workflows around an ontology, with deployment consultations and joint engineering support.
Anthropic and Theorem report AI-written proofs for open-source fuzz targets
Claude produced machine-checked proofs for 239 entry points across 123 libraries, with explicit limits on binaries, inputs, memory and modeled behavior.
OpenAI details offline frontier training defenses after DNS access incident
OpenAI describes layered research containment and confirms an October 1 restart, with earlier DNS and open-source records clarifying the security boundaries.
Anthropic adds physicist Christopher Stubbs to its Long-Term Benefit Trust
Harvard physicist Christopher Stubbs joined Anthropic’s independent Long-Term Benefit Trust, adding scientific experience to the company’s public-benefit governance structure.
Walmart sets out agent shopping profiles and a gateway with spending controls
Walmart announced separate profiles and an early-beta gateway for shopping agents, with customer authorization and spending limits built into the planned experience.
Australian SME study models up to A$93 billion in annual value from wider AI use
Tech Council research estimates the potential value of broader AI adoption, using 2025 adoption data, economic scenarios and earlier workplace studies.
Databricks details AI Runtime GA for managed GPU model training
Databricks outlines managed GPU training, multinode workloads and integrated debugging, with rollout conditions and separate Beta status for custom containers.
Nuix launches Neo AI Agent with source-linked answers and an audit trail
Nuix introduced a conversational investigation agent that uses platform tools, follows data relationships and records its work for professional review.
PHP patches critical PHP2Shell flaws exposed through AI-assisted research
PHP released fixes for two critical memory-safety flaws after AI-assisted investigations by Calif, OpenAI and Wiz, as providers reported attempted attacks.
AI adoption reaches 243 million people in Southeast Asia as digital economy report projects growth
Google, Temasek and Bain estimate broad AI adoption across Southeast Asia, while their new report separates expanding usage from the economics of deeper business deployment.
NVIDIA expands Agent Toolkit with a foundation for enterprise knowledge and predictions
NVIDIA added Agent Knowledge Foundation to its agent stack, combining retrieval, structured-data predictions and business relationships with existing runtime controls.
Mistral and NVIDIA plan AI engineering workflows with shared physics benchmarks
Mistral and NVIDIA will connect physics simulation, model development and design, with planned open datasets, engineering benchmarks and Omniverse integration.
OpenAI plans Google Cloud rollout as selected enterprises enter private preview
OpenAI models and Codex workflows are heading to Google Cloud, with limited private previews, Marketplace purchasing and broader availability planned in the coming weeks.
Mistral launches Voxtral 3 speech models and a toolkit for voice agents
Mistral introduced Voxtral 3 and a voice-agent SDK combining transcription, reasoning, tool calls and speech, with separate models for recorded and live audio.
Google outlines broader AMIE trials as medical AI moves into patient intake
Google researchers detailed plans for larger AMIE studies after supervised patient-intake research, while distinguishing clinical evidence from simulated video consultations.
Microsoft expands sovereign cloud plans for local AI and operational control
Microsoft sets Ignite milestones for local AI and development tools while adding public cloud oversight and a preview for switching to disconnected operations.
Oracle and AstraZeneca set out AI clinical research partnership
A non-binding agreement will explore AI and real-world data for clinical research, beginning with a hypertension programme with M42 in Abu Dhabi.
Google brings Gemini’s business tools into a dedicated small-business hub
Gemini’s new business area combines app connections, reusable brand styles and financial visuals, with a global rollout beginning in the coming days.
Trump names Jay Clayton to lead federal “Super Intelligence Force” as Washington reshapes AI oversight
President Donald Trump named national intelligence director Jay Clayton to lead a new federal AI task force that will coordinate government work on advanced systems.
Nvidia-backed Reflection prepares open-weight model as Western labs try to answer China’s low-cost AI challenge
Reflection is preparing a powerful open-weight AI model that could give developers a cheaper alternative to closed frontier systems from OpenAI, Anthropic and Google.
Trump defends AI data centers in Ohio as energy costs turn compute buildouts into a campaign issue
Trump defended data centers that power AI projects during an Ohio rally, highlighting how compute infrastructure has become a political fight over electricity, land and water.
Apple tightens macOS Full Disk Access controls as AI agents turn privacy permissions into a new safety problem
Apple says it will add clearer controls around macOS Full Disk Access after complaints that AI agents can request unusually broad access to personal data.