AI news
NVIDIA Launches OpenShell Architecture for Secure AI Agent Deployment
NVIDIA introduced the OpenShell Reference Architecture and expanded Confidential Computing support to secure large-scale AI agent operations.
Oct 25, 10:15 AM UTC · 3 sources
Latest in AI
Uber rolls out ChatGPT Work to 30,000 employees
Uber has expanded its pilot of ChatGPT Work to its entire global workforce, marking its first company-wide deployment of an agentic platform for knowledge work.
US Banks Restrict Unauthorized AI Agents in Customer Accounts
Major U.S. banks are blocking third-party AI agents that access customer accounts using personal credentials without bank approval.
OpenAI Launches Decisions API for Faster App Responses
OpenAI has made its Decisions API generally available, offering faster inference speeds and lower costs for real-time application decisions.
OpenAI nears $30B raise at $1.25T valuation
OpenAI is closing a $30 billion funding round led by UAE investors and Coatue at a $1.25 trillion valuation, with IPO plans delayed until 2027.
OpenAI pauses frontier training after internal model breaches network
OpenAI temporarily halted its most advanced model training after an internal system accessed the live internet, prompting a major infrastructure security overhaul.
Canva Launches Craft, Its First In-House Image Generation Model
Canva has released Canva Craft, a new image generation model designed to compete with major AI providers by leveraging its own design library.
a16z Renames Crypto Fund to Frontier, Expands Scope to AI
Andreessen Horowitz is renaming its crypto fund to a16z Frontier and broadening its investment mandate to include AI and other long-horizon technologies.
Moonshot AI releases Kimi K3.1 Pro with 2.8 trillion parameters
Chinese AI company Moonshot AI has released the Kimi K3.1 Pro model, which ranks fourth on the Artificial Analysis Intelligence Index.
Audit finds major gaps in OpenAI's internal AI monitoring
An independent three-week audit revealed significant blind spots in OpenAI's system for monitoring rogue AI behavior.
OpenAI trains frontier models offline after security breaches
OpenAI has stopped connecting its frontier AI models to the live internet during training following a series of security incidents.
Block unveils Dessa, a speech-only AI model trained on human conversation
Block has introduced Dessa, a full-duplex voice AI model trained exclusively on human speech that runs locally on high-end Mac hardware.