Stay ahead on AI.
All in ONE place.

OpenAI has launched "Businesses in ChatGPT" in beta, giving companies a new way to connect with the 1.2 billion weekly ChatGPT users. Through Business Hub, businesses can claim and verify their profile, update product feeds with pricing and availability, and access aggregated engagement insights like mentions and website clicks. Eligible U.S. ecommerce and online service businesses can also build Business Agents — branded AI assistants that answer customer questions within ChatGPT. Shopify merchants can connect directly via Business Hub or Shopify admin.

Anthropic, PBC has launched the roadshow for its proposed IPO of Class A common stock, offering 825 million shares priced between $70 and $78 each. Of those, 801 million shares are offered by Anthropic itself, with roughly 23.6 million sold by existing stockholders. Anthropic will not receive proceeds from the selling stockholders' shares. The company has applied to list on the Nasdaq Global Select Market under the ticker "ANT," with Goldman Sachs and Morgan Stanley serving as underwriters.

NVIDIA has released Alpamayo 2 Edge, a billion-parameter vision-language-action model built for real-time autonomous vehicle reasoning on the DRIVE AGX Thor onboard system. Based on a Nemotron reasoning backbone from Cosmos 3 Edge, it uses surround-camera inputs and Chain-of-Causation reasoning to generate driving trajectories within 160–190ms latency. Post-trained with closed-loop reinforcement learning in AlpaGym, it improves closed-loop driving performance by 48.9%. Partner Embedl achieved 10Hz real-time planning on AGX Thor within 48 hours, while Niulinx reduced displacement error by 44.2% after fine-tuning.

Nvidia CEO Jensen Huang's yearslong relationship with Alice Huang, identified by U.S. officials as a Chinese intelligence asset, facilitated at least $7.5 billion in Nvidia chip sales to China. Alice Huang leveraged her access to Huang to build a corporate network including Shanghai Century Litong and Singapore's Megaspeed, supplying banned H100 and H200 chips to Alibaba, Tencent, and Chinese AI labs. Nvidia's board has launched an investigation, and Jensen Huang has been called to a closed-door congressional briefing on national security concerns.

Nuance Labs has launched Emma, its first AI persona designed to mimic natural face-to-face conversation using facial expressions, not just voice. Emma nods, smiles, and raises an eyebrow at rates similar to human listeners, and her reply timing matches real people 89% of the time across 1,785 test questions. Unlike most AI, Emma will disagree with users and has distinct opinions. Known limitations include occasional mistimed interruptions and lip-sync glitches on poor connections. Emma is available now via API.

Meta has launched Muse Ember, its smallest and most efficient multimodal AI model, designed for high-volume agentic tasks including coding subtasks, tool calls, and document analysis. Available in Muse Code and via Meta Model API, it supports text, image, audio, and video inputs. Priced at $0.10 input and $0.50 output per million tokens, it scores 36 on the Artificial Analysis Intelligence Index and outperforms Claude Haiku 5.5 and GPT-6 Luna on AutomationBench-AA and Terminal-Bench 4.0.

OpenAI has deployed Control Room, a centralized safety system that can halt rogue AI model workloads across its research infrastructure within 35 seconds of a decision to intervene. The system integrates with every research inference workload, including agentic sampling, model training, and evaluation jobs. It requires all agentic workloads to have an approved safety case and operates at the inference layer for reliability. Critically, it fails closed, meaning if Control Room itself goes down, all inference requests are denied until it recovers.

Anthropic's Claude Fable 5.2 successfully designed three working electronic prototypes from scratch in collaboration with Diode Computers: a USB speakerphone, a voice-assistant hub, and a handheld virtual pet. With minimal prompting, Claude wrote specifications, drew schematics, laid out circuit boards, designed 3D-printable enclosures, wrote firmware, and prepared parts lists. Human engineers built the prototypes without any circuitry fixes. Fable 5.2 scored 69% on Diode's PCB Gym benchmark, up from 25% for Fable 5, and 83.5% on BenchCAD 2.1, nearly double its predecessor.

ChatGPT Enterprise is now available on AWS Marketplace, giving organizations a simpler way to discover, buy, and manage the product using existing AWS purchasing relationships and apply eligible spend toward AWS commitments. AWS customers receive the same models, features, and functionality as direct OpenAI customers, with new features released simultaneously. Eligible customers can also access ChatGPT for Healthcare and ChatGPT for Financial Services. The listing complements OpenAI models already available in Amazon Bedrock.