Anthropic files for IPO, the EU enforces its AI Act for the first time, Apple gives Siri memory, and India drafts its first AI safety rules.
Anthropic files for IPO — and reveals it lost $80 billion in the first half of 2026
Anthropic, the maker of Claude, filed its S-1 document with the SEC, beginning the process of going public. The filing reveals $80.7 billion in losses in just the first six months of this year. Goldman Sachs and Morgan Stanley are leading the offering. If the IPO proceeds, Anthropic would be the first major AI lab to trade on a public stock exchange.
Why it matters: This is a landmark moment for the AI industry. Public markets will now force transparency into a sector that has operated largely in private. The losses show just how expensive the AI race has become — and who gets to stay in it.
OpenAI launches GPT-6.1 Terra — a cheaper, faster model for everyday use
OpenAI released GPT-6.1 Terra, a new mid-tier model priced at $1 per million input tokens and $5 per million output, roughly half the cost of its flagship GPT-6.1 Sol. Terra is aimed at developers and businesses that want capable AI without paying premium prices for every query.
Why it matters: More affordable models mean more builders can ship AI-powered apps without burning through budgets. If you are a student or small business owner exploring AI tools, this price drop matters for what becomes available to you.
EU starts enforcing the AI Act — OpenAI and Anthropic get the first formal notices
The European Commission's AI Office sent formal information requests to both OpenAI and Anthropic, marking the first real enforcement action under the EU AI Act. This is the law that requires the most powerful AI models to meet safety, transparency, and risk-management standards. The requests are essentially the EU asking: show us what your most powerful models actually do.
Why it matters: For years AI regulation was all talk. This is the first time a major government body has formally demanded answers from AI labs under law. How companies respond will shape how regulators everywhere proceed.
Apple's iOS 27.2 gives Siri a memory — but not in the EU or China
Apple released iOS 27.2, bringing contextual memory to Siri. This means Siri can now remember past conversations and use that context to give more relevant responses — something rival AI assistants already do. The feature is rolling out globally, except in the European Union and mainland China, where different regulations apply.
Why it matters: Memory is what transforms an AI assistant from a search box into something genuinely useful. This brings Siri closer to what ChatGPT and Google's Gemini already offer. If you are in India, you get the full feature now.
India drafting its first AI safety law — deepfakes and safety rules expected by November
India's IT Ministry is preparing a consultation paper on AI safety and deepfakes, expected within weeks. It would be India's first structured legal framework addressing AI-generated content, safety standards for powerful models, and accountability rules. The paper is expected before November 2026, giving stakeholders time to respond before any rules are finalised.
Why it matters: India has been watching deepfake harms grow — especially affecting women — without a clear legal tool to respond. This is the government signalling it is ready to act. If you build or use AI products in India, this consultation is your chance to shape what those rules look like.
UK bets £100M on an AI biosecurity lab — with OpenAI, Anthropic, and DeepMind inside
The UK government announced a £100 million AI Biosecurity Lab, with OpenAI, Anthropic, and Google DeepMind as founding partners. The labs will focus on using AI to detect biological threats and prepare defences. The three companies are also contributing up to £200 million in model credits and compute access on top of government funding.
Why it matters: The biggest AI labs are now jointly building tools to protect against AI misuse in one of the highest-risk domains. It is also a signal that governments want AI companies at the table when critical national security infrastructure is being built.
The EU deepfake ban takes effect December 2 — every platform has to be ready
A new EU regulation banning AI systems that generate non-consensual intimate imagery comes into force on December 2, 2026 — just over five weeks away. Platforms operating in Europe that have not yet built removal pipelines or compliance systems are now on a tight clock. This is separate from the AI Act enforcement notices sent today. The two together signal that 2026 is the year European AI rules stopped being theoretical.
India's AI startup funding hits $1.1B+ in 2026 + Pakistan targets election deepfakes
Indian AI startups raised over $1.1 billion in the first nine months of 2026, a 265% jump year-on-year, with agents and infrastructure tools attracting the most capital. Separately, Pakistan's Election Commission has published draft rules to ban AI-generated deepfakes in election campaigns — one of the first such moves in South Asia.