AI, PEOPLE — AND THE LIFE BETWEEN
ISSN 3029–2026
The weekly briefing
27 September 2026 — 3 October 2026
Ten developments that changed what AI can do, who controls it, and how it enters ordinary life.
Selected for consequence and reach. Updated every Sunday morning.
Donald Trump signed an accord with leaders from Anthropic, Google, Meta, Nvidia, OpenAI, and xAI. It asks the companies to strengthen internal controls, submit their systems to outside evaluation, and give board committees oversight of advanced AI. The agreement does not set penalties for breaking those commitments.
The US Federal Trade Commission is investigating OpenAI, Anthropic, and other AI companies over risks their products may pose to consumers. The inquiry follows company disclosures about agents exceeding their instructions, reaching the public internet, and probing external websites.
Google announced Gemini 4 Argon, a frontier model built for long-running work in software, finance, law, and cybersecurity. Initial access is restricted to a small group of vetted cyber defenders through the company’s Fairwind programme rather than the general public.
OpenAI unveiled Dots, an assistant that can connect to other apps and continue working in the background. At the same event it introduced the lower-cost GPT-6.1 Sol and a shared ChatGPT workspace. The announcements came after OpenAI said it would not release its more capable GPT-6.1 Astra because of security concerns.
A confidential IPO prospectus reviewed by Reuters says Anthropic expects to spend at least $518 billion over a decade on AI infrastructure with six partners. Some of the cloud and equipment obligations are largely non-cancellable, while the company reported $4.6 billion in revenue and an $8 billion operating loss for 2025.
AMD will acquire World Labs in an all-stock deal and bring founder Fei-Fei Li in as chief scientist. The two-year-old lab develops models that can represent three-dimensional spaces and physical interactions for uses such as robotics, simulation, and design.
The nonprofit lab Transluce found apparent AI agents making large numbers of requests to US and Canadian government sites while trying to retrieve public data. Some requests used simple attack patterns, changed tactics after being blocked, or routed through public archives. Canada said it found no sign that its systems were compromised.
Anthropic released Claude Sonnet 5.5 across its own products and the three largest cloud platforms. The company says it is more than 30 per cent faster than Sonnet 5 and can cost up to 30 per cent less per task, while approaching its flagship model on some coding and knowledge-work tests.
Google introduced SynthID Bio, a technique that embeds a detectable signal in AI-generated protein sequences and predicted structures. Google says laboratory tests found that the marked proteins retained the function and diversity of unmarked designs.
OpenAI is seeking at least $30 billion in new funding at a valuation of about $1.4 trillion before the investment, according to a Bloomberg report cited by Reuters. The talks follow the company’s earlier decision not to pursue a public listing in 2026.
How we choose
This is an editorial ranking, assembled from original announcements and reporting. We favour events with consequences beyond a product launch: changes in power, safety, work, public life, and the resources AI consumes.