🏢 OpenAI creates new unit with $4 billion investment to aid corporate AI push ↗
OpenAI is setting up a new majority-owned deployment company, backed by more than $4 billion, to help businesses put AI into their operations - not just show it off in a glossy meeting-room demo.
It is also acquiring Tomoro, an AI consulting firm with around 150 engineers and deployment specialists. The aim is direct: place frontier-AI people inside companies and identify where the technology can make the largest dent. Very “AI meets Big Consulting in a tailored suit.” (Reuters)
🛡️ AI labs should pass safety review to get US government contracts, group says ↗
A policy group urged U.S. officials to require frontier AI labs to pass security reviews before they can win major government contracts.
The proposal focuses on risks such as cyberattack assistance and weapons-development capabilities, and would apply to larger AI companies based on compute spend or AI revenue. It is a carrot-and-stick arrangement, essentially - access to federal money, but only after the model gets checked under the bonnet. (Reuters)
🇪🇺 EU says OpenAI offers to open access to cybersecurity model, Anthropic not there yet ↗
The European Commission welcomed OpenAI’s offer to provide access to its cybersecurity features, while saying talks with Anthropic had not reached that stage.
OpenAI framed the move as part of an EU cyber action plan, aimed at giving trusted actors defensive tools for shared security and public safety. It sounds cooperative, and it is also a canny bit of regulatory weatherproofing. (Reuters)
⚡ OpenAI just released its answer to Claude Mythos ↗
OpenAI launched Daybreak, a security-focused AI initiative built around detecting and patching vulnerabilities before attackers get there first.
The system combines GPT-5.5-Cyber with Codex Security, using an organization’s code to model likely attack paths, validate vulnerabilities, and automate detection of higher-risk issues. Slightly superhero, slightly fire-alarm-in-a-server-room. (The Verge)
🗣️ Thinking Machines wants to build an AI that listens while it talks ↗
Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, announced “interaction models” - AI designed to process input and generate responses at the same time.
The idea is full-duplex conversation: less turn-taking chatbot, more phone call. Its small research model reportedly responds in 0.40 seconds, but it is still a preview, not a public product. Promising? Yep. Proven in the wild? Not quite. (TechCrunch)
📰 Digg tries again, this time as an AI news aggregator ↗
Digg is back again, improbably enough, and the new version is starting as an AI-news ranking site rather than another Reddit-style community.
The reboot is meant to track influential voices and surface stories worth paying attention to. It is still described as raw and buggy, which feels refreshingly candid - like launching a kite in a server rack. (TechCrunch)
FAQ
Why is OpenAI creating a new corporate AI unit?
OpenAI is creating a majority-owned deployment company to help businesses bring AI directly into their operations. The emphasis is not just on demos, but on identifying practical areas where frontier AI can make a measurable difference. The acquisition of Tomoro adds engineers and deployment specialists who can work inside companies on existing workflows.
What does the proposed AI safety review for government contracts mean?
The proposal would require frontier AI labs to pass security reviews before receiving major U.S. government contracts. It is designed to reduce risks from advanced models, including cyberattack assistance or weapons-related capabilities. In practice, it connects access to federal money with stronger checks on how powerful AI systems are tested, secured, and controlled.
How is OpenAI using AI for cybersecurity?
OpenAI’s Daybreak initiative is described as a security-focused AI effort for finding and patching vulnerabilities earlier. It combines cybersecurity models with tools that can inspect an organization’s code, map likely attack paths, validate vulnerabilities, and support automated detection. The goal is defensive: helping trusted teams identify serious issues before attackers exploit them.
What are interaction models in AI?
Interaction models are AI systems designed to process input and generate responses at the same time. Thinking Machines Lab presents this as a step toward more natural, full-duplex conversation, closer to a phone call than a turn-based chatbot. The article notes that the model is still a preview, so it should be treated as promising research rather than a finished public product.
Why is Digg returning as an AI news aggregator?
Digg is returning as an AI news ranking site rather than immediately trying to rebuild a full Reddit-style community. The new version aims to track influential voices and surface stories worthy of attention. According to the article, it is still early and somewhat raw, but the direction is clearly tied to AI-powered news discovery.