OpenAI is heading for a $730B IPO, 100+ companies sound the alarm on AI cyberattacks, and the EU's enforcement teeth are now fully sharp: September 3 is a day of reckoning.
OpenAI is preparing a confidential IPO filing, valued at $730 billion
OpenAI is preparing to file confidentially for an initial public offering in the coming weeks, with a potential market debut as soon as September 2026. Private markets already value the company at $730 billion, making it one of the most anticipated listings in tech history. This follows Anthropic crossing $30 billion in annualised revenue and breaking 1,000 enterprise customers spending over $1 million a year.
Why it matters: An OpenAI IPO would move AI from a venture-backed experiment to a publicly traded industry. It also forces every company still treating AI as optional to reckon with its scale: OpenAI is bigger than most banks.
100+ companies (including OpenAI, Anthropic, and Google) sign rogue AI defense letter
Over a hundred technology companies, including OpenAI, Anthropic, Google, and Microsoft, have signed an open letter urging governments and private sectors to coordinate defences against AI-driven cyberattacks. The letter coincides with OpenAI's Astra model becoming the first AI system to exceed the Preparedness Framework's "Critical" cybersecurity threshold, and autonomously finding two real zero-day vulnerabilities in security tests.
Why it matters: The same organisations building powerful AI are now formally warning that it can be used against us. This is the clearest signal yet that AI safety isn't just a philosophical debate, it is an active infrastructure concern.
EU AI Act: enforcement is now fully live, with real fines for the first time
The EU AI Act's full enforcement powers came into effect on August 2, 2026, and those rules are now active across Europe. High-risk AI systems must carry conformity assessments and CE markings. AI-generated content must carry machine-readable watermarks. The EU AI Office now has the authority to impose massive fines on general-purpose AI providers, fines that could not be levied until this month. Any app or service that uses AI and operates in Europe is now operating under binding law.
Why it matters: Until August, the EU AI Act had obligations but limited enforcement. Now it has teeth. If you use AI tools in a European context (or if your employer does) this shift from guidelines to fines changes the conversation entirely.
India is building an AI payments system: AI agents could soon spend your money on your behalf
India is preparing a framework that would let AI agents make small, repetitive payments through UPI (India's ubiquitous digital payments system), without requiring user approval for every transaction. The system would include spending limits, identity verification, and predefined rules. The proposal is expected to be presented at the Global Fintech Fest in Mumbai, where regulators and banks will debate guardrails.
Why it matters: This is the first formal move by a major government to embed AI agency into a national payments infrastructure. If it launches, India becomes a live test case for what it means to give an AI model a wallet, and the rules it sets will influence how the world thinks about autonomous AI spending.
New models land this week: Fable 5.1, Qwen3.8 27B, and Gemini 3.8 Flash
Anthropic's Fable 5.1 became generally available on September 1. Alibaba's Qwen3.8 27B launched on September 2. Google's Gemini 3.8 Flash is now live. Behind the scenes, Dell reported $16.4 billion in AI server revenue for a single quarter (up 58% year on year) with a $95 billion backlog of AI orders. The infrastructure race and the model race are running in parallel, and both are accelerating.
Why it matters: More capable models at lower costs, plus more computing muscle being built to run them. If you've been waiting for AI to get cheap enough or good enough to use in your work, the window is now.
What happens when AI companies go public
OpenAI's IPO would be a turning point, not just for the company, but for how the whole AI industry is governed. Public companies answer to shareholders, not just mission statements. OpenAI's non-profit origins and safety commitments will face new pressure the moment quarterly earnings reports begin. Watch whether the IPO filing changes anything about how the company talks about safety, speed, and the commercial trade-offs it is willing to make. The answer will tell us a lot about the next decade of AI.
6 million teachers now use AI daily + AI funding in 2026 already broke last year's record
MagicSchool AI has crossed 6 million educators signed up across 160 countries, the largest adoption of any teacher-facing AI tool. Separately, global AI startup funding in the first half of 2026 reached $510 billion, already exceeding all of 2025's $440 billion. The money is concentrated: 88% went to US companies, and Tripo AI raised $446 million alone for 3D generative AI.