AI news, 12 July: OpenAI and SpaceXAI race to release rival flagship models

OpenAI and SpaceXAI launch rival flagship models within a day of each other. OpenAI publicly rolled out its GPT-5.6 family — three versions called Sol, Terra and Luna — on 9 July, ending a roughly two-week government-coordinated preview period. According to the AI news tracker unrot.co, Terra is designed to match the performance of OpenAI's previous flagship at roughly half the running cost, while Sol targets top-end coding tasks and Luna focuses on speed and low price. The very same week, SpaceXAI — the company formed when SpaceX merged with Elon Musk's xAI earlier this year — released its own Grok 4.5 model, co-developed with the coding tool Cursor. The back-to-back launches show how fast major labs are now trading blows on price and performance, though the specific benchmark claims are not independently verified.

Google's next big model is delayed again, reportedly for a good reason. Gemini 3.5 Pro, once expected earlier this year, is now said to be heading for a public launch around 17 July, according to unrot.co, which cites AI community sources. The outlet reports that Google DeepMind scrapped its earlier plan to fine-tune the existing 2.5 Pro model and instead started training a new model from scratch, a much bigger undertaking that could explain the repeated delays. This remains an unconfirmed leak rather than an official Google announcement.

Chipmaker SK Hynix's Nasdaq debut highlights the AI hardware boom. South Korean memory-chip maker SK Hynix began trading on the Nasdaq exchange this week under the ticker SKHY, in what unrot.co described as one of the largest such listings on record. It reflects continued investor enthusiasm for firms supplying memory and processing chips for AI data centres.

UN meeting warns that AI safeguards are not keeping pace. At the UN Global Dialogue on AI Governance held in Geneva earlier this week, government officials, researchers and civil-society figures discussed how to regulate a technology moving faster than the rules meant to contain it, according to UN News. Speaking at the event, AI researcher Yoshua Bengio said AI "is approaching or surpassing human capabilities in many domains," warning it is outpacing scientific understanding and governments' ability to adapt. Journalist Maria Ressa separately raised alarm about AI-driven disinformation, calling the risk to democratic debate an "information Armageddon." The panel's newly published report also flagged that millions of children are already using AI tools without adequate safeguards in place.

New EU rule requires driver-attention AI in every new car. Since 7 July, every newly registered car in the European Union has been required to include a driver-distraction detection system that uses AI to analyse a driver's gaze and head movements without recording or transmitting footage, according to a weekly AI news roundup. The rule marks one of the more tangible ways AI regulation is now reaching ordinary consumers, folded quietly into broader EU vehicle-safety law.