Anthropic opens its books with revenue up 14x, OpenAI prepares a post-midterm policy blitz, and India breaks into the top 3 of the global AI vibrancy index.
Anthropic opens its books: revenue up 14x, IPO in sight
Anthropic published its financials ahead of a planned November IPO. Revenue has grown roughly 14 times over the past year. The headline loss of $80.7 billion is almost entirely an accounting charge, not cash spent. The company was last valued at $965 billion in its May 2026 Series H round.
Why it matters: Anthropic going public would be the largest AI company listing in history. It signals that the era of AI companies operating quietly as private labs is ending — and that public markets are about to have a direct stake in how AI develops.
OpenAI is heading to Washington with a plan for AI "speed limits"
OpenAI's chief global affairs officer Chris Lehane is launching a coordinated policy and PR campaign starting November 9, timed to the post-midterm lame-duck session in Congress. The centrepiece is mandatory "speed limits" — caps on how fast the most powerful AI systems can be deployed or scaled.
Why it matters: This is OpenAI voluntarily asking for its own industry to be regulated. Whether you read that as genuine responsibility or a strategic move to lock out competitors, it means AI policy is about to move faster in the US. Whatever rules land here will shape what the rest of the world does next.
Google's Gemini 4 Carbon: a cybersecurity AI with restricted access
Google DeepMind launched Gemini 4 Carbon, a model purpose-built for cybersecurity work. Early access is restricted to vetted cyber defenders through its Fairwind programme. The model is the second in the Gemini 4 family after Argon, which launched September 30 and went first to the US government.
Why it matters: The biggest AI labs are now building models specifically for security professionals, not just general users. Access is deliberately limited — which is a new kind of AI gatekeeping worth watching as a pattern.
FTC begins enforcing Take It Down Act on sexual deepfakes
The US Federal Trade Commission has started actively enforcing the Take It Down Act, which requires platforms to remove non-consensual AI-generated sexual imagery. This shifts deepfake harm from a policy debate into a compliance obligation for major platforms.
Why it matters: This is the first time a major country has moved from passing a law about deepfake harm to actually enforcing it with a federal agency. India is watching closely — its own AI regulation consultation paper is expected within days.
India ranks 3rd globally in AI vibrancy. Eight new LLM startups funded.
India has climbed to third place in the global AI vibrancy index, behind the US and China. Separately, the IndiaAI Mission has selected eight more startups — including Fractal Analytics and BharatGen — to build indigenous large and small language models. The government's Rs 2,000 crore startup fund is now actively disbursing.
Why it matters: India's AI ambitions are no longer just aspirational — the infrastructure, funding, and recognition are all moving in the same direction at the same time. If you're a student or professional building in AI in India right now, the window is open.
Perplexity enters language learning, taking on Duolingo directly
Perplexity has launched an AI-powered language learning feature, entering a market where Duolingo recently pivoted to an AI-first model. UCL and Google DeepMind have also co-developed an AI Research Foundations curriculum for university students and early-career researchers — one of the first partnerships of this kind between a frontier lab and a major university.
Why it matters: AI tools built for learning are arriving faster than institutions can evaluate them. Students who learn to use them well get a compounding advantage. Those who wait for their school to assign one may be a year behind.
The post-midterm window for AI regulation in the US
US midterms just closed, and the lame-duck congressional session opens. OpenAI has planned its policy campaign to land precisely in this window. The bills and frameworks that get introduced or advanced now could set the baseline for global AI governance for the rest of the decade. Whatever passes — or fails — here will ripple quickly into policy debates in India, the EU, and across Southeast Asia.
GPT-6.1 Terra cuts costs by half + Claude Haiku 5.5 adds effort controls
OpenAI's GPT-6.1 Terra is now available at half the per-token cost of its predecessor, making the frontier more accessible to developers on a budget. Anthropic's Claude Haiku 5.5 is the first Haiku model to include effort controls and a one-million-token context window — useful for document-heavy professional workflows without the cost of the full Opus tier.