Happy Friday! Last week ChatGPT and Gemini each hit 1 billion users as OpenAI also launched specialized cybersecurity AI and Meta released a 30B open-source model for PCs. Anthropic announced watermarking for all Claude-generated text.

GENERATIVE AI MILESTONE

ChatGPT and Gemini each reach 1 billion users

Google CEO Sundar Pichai announced that Gemini has reached 1 billion monthly users, making it Google's fastest-growing product ever and the 14th Google product to hit that milestone. However, Google is not the first AI platform to reach this mark. OpenAI's ChatGPT also crossed the 1 billion user threshold, making both leading AI assistants members of an exclusive club.

AI STARTUP FUNDING

River AI Raises $1.1B for Model Customization

River AI, a startup specializing in enterprise open-source AI model customization, has raised $1.1 billion across a seed and Series A round. Backers include General Catalyst, AMP PBC, Nvidia, AMD Ventures, Y Combinator, and Temasek. The company is led by Igor Babuschkin, a co-founder of xAI and former DeepMind researcher. River AI's flagship product, the River API, uses a technique called LoRA to customize large language models ranging from 35 billion to 1 trillion parameters in 15 to 20 minutes. The company claims its models can be four times more cost-efficient than proprietary alternatives and plans to expand into agent personalization and custom AI chip development.

NEW GENERATIVE AI LAUNCH

OpenAI Launches Specialized AI for Advanced Cybersecurity Tasks

OpenAI has released GPT-5.6-Cyber, a fine-tuned model built for vulnerability research and exploit development, achieving 95% completion on advanced cybersecurity tasks. It is designed exclusively for approved defenders and reduces refusals on dual-use cybersecurity requests. The model is based on GPT-5.6 Sol and trained to handle high-risk tasks such as zero-day vulnerability discovery, exploit chain development, authentication bypass, and privilege escalation. Performance is measured using OpenAI's internal Advanced Cybersecurity Completion Rate benchmark.

NEW GENERATIVE AI LAUNCH

Meta Launches 30B Open-Source AI Model for PCs

Meta released Muse Glimmer, a 30-billion-parameter open-source language model optimized to run on personal computers and Macs using a single consumer-grade GPU. Through quantization and other techniques, Meta reduced its memory footprint from 55GB to under 20GB. The model uses speculative decoding for faster responses and was trained to retry failed tasks autonomously. Meta CEO Mark Zuckerberg accompanied the launch with an essay calling for reduced open-source AI regulations and urging labs to grant governments early model access to address cybersecurity risks. Muse Glimmer outperformed comparable models on half of 24 benchmarks tested, covering coding, research, and scientific analysis.

NEW GENERATIVE AI LAUNCH

Anthropic to watermark all Claude-generated text

Anthropic will watermark text generated by its AI models, including Claude, to comply with the EU AI Act's Transparency Code, which took effect August 2. The company confirmed all models released after that date will automatically watermark AI-generated text and files, using the C2PA open standard for files. The watermark travels with copied and pasted text and applies across all Claude products and surfaces. Anthropic joins Google, Meta, Microsoft, OpenAI, and others in committing to the EU's transparency code, as platforms face growing pressure to label AI-generated content amid regulatory scrutiny and user backlash.

AI STARTUP FUNDING

Skan AI Raises $63M to Map Enterprise Work

Skan AI has closed a $63 million Series C round to expand its process intelligence platform, which uses on-device screenshot analysis to map how enterprise work gets done without storing sensitive data. The anonymized metadata feeds AI agents modeled on top-performer behavior. The round was co-led by Cathay Innovation and Dell Technologies Capital, bringing total funding to $120 million. Customers include a quarter of the Fortune 50 and seven of the 10 largest U.S. banks. The company reports 300% year-over-year revenue growth and claims more than $500 million in cumulative customer savings.

NEW AI SIGN LANGUAGE MODEL

Google AI Model Translates Sign Language to Text

Google DeepMind has launched SL2T, a sign-language-to-text AI model debuting on the Pixel 11 smartphone. The tool powers a new sign-to-text feature in Gboard and Live Transcribe, initially supporting American Sign Language to English translation. Google calls it the first such model available in a real-world consumer product. Trained on over 100,000 hours of sign language data across 50 languages, SL2T uses on-device computer vision to track body movements without uploading video footage. Google has also established an AI Sign Language Advisory Committee with deaf organizations to guide responsible deployment and plans to expand support to additional sign languages.

AI INFRASTRUCTURE DEAL

IBM Signs $240M Deal With Together AI

IBM will provide infrastructure to AI startup Together AI under a $240 million partnership announced August 11. The deal follows Together AI's recent $800 million funding round, which included backing from Nvidia, whose chips IBM will use to power the workloads. IBM will deploy a large cluster of HGX B300 systems featuring Nvidia Blackwell Ultra chips, available through IBM Cloud in the first half of 2027. Together AI will use the hardware to run open-source AI inference workloads for enterprise customers on its public cloud platform, which currently processes around 400 trillion tokens per month.

AI INVESTMENT STRATEGY

Jensen Huang Turns AI Chips Into Wall Street Assets

NVIDIA CEO Jensen Huang is reframing the company's chips as revenue-generating assets rather than mere hardware, drawing Wall Street deeper into the AI infrastructure race. The vision is tied to a $500 billion AI investment outlook. As AI demand accelerates, Huang's pitch positions compute power as a financial instrument capable of generating returns, potentially reshaping how investors and institutions think about AI infrastructure spending.

AI CYBERSECURITY PARTNERSHIP

Palo Alto Deploys OpenAI Cyber Models in Customer Networks

Palo Alto Networks' Unit 42 consulting arm is deploying OpenAI's frontier cyber models inside customer environments through an expanded service called Frontier AI Exposure Analysis. The models scan for vulnerabilities, misconfigurations, leaked credentials and unmanaged attack surfaces, with adversary simulation used to confirm whether exposures are genuinely exploitable. The service uses OpenAI's Daybreak Red program, built on GPT-5.6-Cyber, which completed 95% of advanced cybersecurity tasks in internal benchmarks. Unit 42 consultants oversee the models and validate findings before delivering ranked remediation plans. Palo Alto also holds a separate partnership with Anthropic through Project Glasswing.

Upcoming Data & AI Conferences

NVIDIA GTC Berlin 2026

Tue, Oct 20 - Thu, Oct 22, 2026 | Berlin, Germany

Curated with AI by EMC2 AI

Brought to you by Estevan McCalley

Share this issue

Follow Estevan

Like this? Subscribe for the weekly AI roundup.