Two new Claude models, Apple's first paid AI tier, AI deepfakes flooding US midterms, and India's national AI rules finally arriving — a lot moved today.
Anthropic ships Claude Fable 5.2 for everyone — and locks Mythos 5.2 behind a vetting process
Anthropic released two models today. Claude Fable 5.2 is generally available to all users and runs about 30% faster than the previous version. Mythos 5.2 is different: it's restricted to vetted organisations in cyber and life sciences only, with Anthropic approving access case by case. The company described it as one of its most capable models to date.
Why it matters: This is the clearest split yet between AI for everyone and AI that's too powerful to release openly. If you use Claude for work, Fable 5.2 is an upgrade you get automatically. Mythos signals that the frontier is now being rationed — not just priced differently.
Apple charges for AI for the first time — $49.99 for an Intelligence usage pack
Apple launched paid usage packs for Apple Intelligence, priced at $49.99. This is the first time Apple has charged separately for AI features, which were previously free with the operating system. The packs appear to be aimed at power users who hit the limits of free AI usage on their devices. Meta also released Muse Ember, a new lightweight model aimed at agent tasks, and Google shipped Gemini Omni 1.2 Flash with video generation capabilities.
Why it matters: Apple charging for AI is a signal that the free-AI era on consumer devices is ending. If you're a heavy user of AI on your iPhone or Mac, expect to start paying. And the race across Apple, Meta, and Google to ship faster, cheaper models is accelerating.
AI-made political ads flooded the US midterms — and most didn't disclose they used AI
The 2026 US midterms are the first major test of the state deepfake laws, with 33 states having passed laws targeting AI-generated content in political ads. Despite the laws, AI-made political ads ran widely on TV and social media, mostly without any AI disclosure label. Courts have already struck down two state deepfake laws on free speech grounds, leaving enforcement patchy. Brown University researchers published a report arguing the problem is not just the content — it's the entire digital ecosystem behind it.
Why it matters: Laws exist. Enforcement doesn't. This is the gap between AI regulation passing and AI regulation working. What happens in the US this month sets a template — or a warning — for every other democracy heading into an election with AI on the loose.
India's national AI rules are finally coming — MeitY consultation paper expected this month
IT Minister Ashwini Vaishnaw confirmed in October that MeitY would release a consultation paper on AI regulation within a month, putting the draft squarely in November. The paper is expected to cover safety, deepfakes, and startup compliance. At the same time, India's cybersecurity agency CERT-In released AI Security Auditing Guidelines covering model poisoning, prompt injection, and agentic AI threats — a risk-based framework for any organisation in India that develops, buys, or deploys AI systems.
Why it matters: India has been watching and waiting while the EU, US, and China built their AI frameworks. That window is closing. If you work in a startup, a bank, or a large company using AI tools, what comes out of MeitY this month will likely shape your compliance requirements for years.
Bihar becomes the latest Indian state to pass an AI policy — and the courts are protecting celebrities from deepfakes
Bihar's state cabinet approved the Bihar AI Policy 2026 on November 2, aimed at attracting AI companies and startups to the state. It follows similar moves by other Indian states to build local AI ecosystems rather than wait for a national framework. Separately, the Bombay and Delhi High Courts have issued orders protecting celebrities' likenesses from AI deepfakes, and the National Commission for Women is running an AI literacy programme for women across India.
Why it matters: India's AI story is now being written at every level at once — national, state, judicial, and community. The window for Indian professionals and students to get positioned inside this ecosystem is open and moving fast.
Nvidia and a Chinese lab both launched major robotics AI models. And fal just raised $1.3B.
Nvidia unveiled GR00T 2 at the CoRL 2026 robotics conference in Austin — a 2-billion-parameter vision-language-action model for physical robots. The same week, Chinese physical-AI lab Nature Will released Natura-1, a 50-billion-parameter model for dexterous robotic hand control. Meanwhile, AI infrastructure company fal closed a $1.3B funding round at a $16B valuation, one of the largest AI raises of the month.
Why it matters: Physical AI — robots that can actually handle objects and navigate the world — is now where the model race is heading. This isn't science fiction or five years away. The money and the models are both arriving at the same time.
India's regulation window: three bodies, one week, one chance to get it right
MeitY is about to release its national AI consultation paper. CERT-In just published security audit guidelines. Bihar passed a state AI policy. And the courts are issuing their own orders on deepfakes. All in the same week. The question is whether these different pieces point in the same direction, or whether India ends up with a patchwork of rules that contradict each other before a national law is even in place. Watch what the MeitY paper actually says about startup compliance — that's the detail that will tell you how ambitious or cautious the government really is.
Claude can now read your bank account (via Plaid) + NotebookLM is now Gemini Notebook
Anthropic added read-only bank account access to Claude via a Plaid integration, available on Pro, Max, and Ultra plans. Claude won't train on what it reads. Separately, Google renamed NotebookLM to Gemini Notebook — same tool, new branding. Both are small moves, but they signal how deeply AI is embedding into daily financial and research workflows.