The top things worth knowing about in AI today.
The 2026 World AI Conference and High-Level Meeting on Global AI Governance runs in Shanghai from today until 20 July, with President Xi Jinping delivering the opening keynote. It is his first in-person appearance since the event began in 2018, and more than 1,100 companies are exhibiting over 3,000 products, including about 300 global debuts. Beijing is using the event to push its own framework for global AI rules, which matters for any business that will eventually work under competing US, EU and Chinese standards.
Read moreThe Future of Life Institute graded leading labs across six categories in its Summer 2026 AI Safety Index and gave no company higher than a C+. Anthropic led on 2.66, OpenAI and Google DeepMind earned a C, Meta a D+, and xAI, DeepSeek and Mistral failed. The report notes several labs have weakened earlier pledges to pause development at set risk thresholds and reversed bans on military use, which is worth knowing if safety claims factor into your vendor choice.
Read moreThe New York Times, the New York Daily News and 15 other news organisations asked a Manhattan federal judge on 9 July to sanction OpenAI, alleging it withheld and destroyed evidence in their copyright case. The publishers say OpenAI kept deleting ChatGPT conversation logs after a preservation order took effect, and had already assembled a database of roughly 78 million de-identified conversations to assess its own infringement exposure. The ruling will shape how much AI vendors must disclose about what their models were trained on and what they reproduce.
Read moreThe European Commission published its Action Plan on Cybersecurity and AI on 7 July, establishing a dedicated EU capacity to run third-party evaluations of advanced AI models before they can be placed on the European market, targeted to be operational by 2027. The Commission and ENISA will finish a blueprint for structured access to frontier models in the fourth quarter of 2026, alongside a secure testing platform for energy, transport, health, finance and public administration. Vendors selling models into Europe should plan for an external review step rather than self-certification.
Read moreOpenAI released GPT-5.6 and ChatGPT Work on 9 July, combining ChatGPT with Codex so Pro, Enterprise and Edu users can produce documents, presentations and websites from plain-language instructions. GPT-5.6 ships as three models: Sol, Terra and Luna, with Sol reported to be 54% more token efficient on coding tasks. Free and Go users get Terra, Plus subscribers get Sol, and the top effort levels are limited to Pro tiers. The claim is finished output rather than draft text, so the practical test is whether the deck it returns needs less rework than one you would build yourself.
Read moreMicrosoft will invest A$25 billion in new digital infrastructure in Australia by the end of 2029, its largest commitment in the country to date, alongside a memorandum of understanding with the federal government covering energy, water and local jobs. It also pledged to help three million Australians build workforce-ready AI skills by 2028, triple its previous target, and signed a framework agreement with the ACTU. A separate content deal with Nine Entertainment lets Copilot use its journalism with attribution and revenue flowing back to the publisher.
Read moreIn a 14 July speech, Federal Reserve Governor Michael Barr argued the open question is not whether AI raises output but who captures the gains, noting the top fifth of US households already earn 52% of all income. He pointed to a usage gap, with 43% of workers holding graduate degrees using AI against 10% of those with a high school education, while citing research showing AI lifted task completion 40% and quality 18%, with the largest gains going to lower-performing workers. His read is that the tools help the people furthest behind the most, but only where access and training actually reach them.
Read moreA short read every week — the few things worth your time, and nothing that isn’t.