AI news, 10 July: OpenAI's GPT-5.6 lands the same day xAI's Grok 4.5 goes public

OpenAI launches GPT-5.6, hours after xAI's Grok 4.5 goes live. OpenAI began rolling out its newest flagship model family, called GPT-5.6, across ChatGPT, its developer API and its coding tool Codex on 9 July, according to the company's own announcement. Just a day earlier, Elon Musk said his company xAI would make its rival Grok 4.5 model available to the public this week, describing it as "an Opus-class model, but faster... and lower cost." Industry trackers noted that, for a brief window, several major AI labs' newest systems were all publicly available at once, intensifying competition among ChatGPT, Grok and other assistants.

UN summit spotlights AI's "catastrophic harm" risks. Governments, companies and civil-society groups met in Geneva this week for the UN Global Dialogue on AI Governance, a meeting meant to work out international rules for a technology that is evolving faster than the rules meant to contain it, according to UN News. A UN advisory panel flagged that children are already using AI to learn and even seek emotional advice, while safeguards are failing to keep pace with adoption.

Illinois signs sweeping AI safety law. Governor JB Pritzker signed the Artificial Intelligence Safety Measures Act on 6 July, following similar legislation in California and New York; according to Capitol News Illinois, the three states together account for roughly 40% of the US AI market, effectively setting a national standard in the absence of federal rules. The law requires major AI developers to publish plans for identifying "catastrophic risk" and to report dangerous incidents within 72 hours, or within 24 hours in emergencies.

Sam Altman revives idea of a public stake in OpenAI. According to MIT Technology Review, OpenAI chief executive Sam Altman is again discussing giving the US government a stake, reportedly around 5%, in the company — an idea meant to let ordinary Americans share in AI-driven profits while easing worries about job losses. At OpenAI's current valuation, that stake could be worth roughly $320 per US household, though the proposal's details are unclear and it is not yet official policy.

DeepSeek's chip ambitions rattle Nvidia. Chinese AI developer DeepSeek is working on its own chip for running AI models rather than training them, Reuters reported, sending Nvidia shares modestly lower amid investor concern about growing competition in the market for AI "inference" hardware.