Daily AI Brief: November 13, 2026

India launched its own AI models in three languages, New York turned safety talk into law, and open-source video got its biggest release yet.

India releases its own AI models: EKA-1, Dhīr 2B, and Dhīr 5B

The IndiaAI Mission released three foundation models built for multilingual use — EKA-1 and two sizes of the Dhīr model — targeting Indian language speakers who have been underserved by English-dominant global AI. These are publicly available and not locked inside a single company's product.

Why it matters: This is India building its own AI stack rather than depending entirely on US or Chinese models. For students and professionals working in regional languages, this is a direct step toward tools that actually work for them.

Google Gemini 4 Hydrogen is now the default across Search, the Gemini app, and AI Studio

Google rolled out Gemini 4 Hydrogen as the default model in AI Mode across Search, the Gemini app, and Google AI Studio. This follows Gemini 4 Carbon, which went live a few days earlier for developers and AI Ultra subscribers. Two significant Gemini upgrades in less than a week.

Why it matters: The model billions of people interact with daily just changed underneath them. If you use Google Search in AI Mode, you are already on a new model whether you noticed or not.

New York's RAISE Act kicks in: AI labs must now register, report incidents and publish safety plans

New York's Responsible AI Safety and Education Act (RAISE Act) registration requirements began this month for frontier AI developers — defined as companies with more than $500 million in annual revenue. Full obligations take effect January 2027 and include publishing safety frameworks, issuing model transparency reports, and reporting critical safety incidents within 72 hours.

Why it matters: This is the first US state law to put real reporting teeth on big AI labs. It signals that the US is moving toward fragmented state-by-state regulation rather than waiting for a federal framework — which may push companies to meet the strictest bar nationwide.

Indian AI startup funding hits $1.6B in nine months — up 136% year on year

Tracxn data shows Indian AI startups raised $1.6 billion across 81 rounds in the first nine months of 2026, compared to $677 million a year ago. The growth is real, but four companies — Neysa, Sarvam, Yotta, and Emergent — took 72% of the total. Deal count also fell 43%, meaning money is concentrating in fewer bets.

Why it matters: There is serious capital moving into Indian AI, but it is not spread evenly. If you are building in this space, the question is less "is there money?" and more "how do I get in front of the people writing the big cheques?"

Runway releases Praxis-1 video model with open weights under Apache 2.0

Runway released Praxis-1, a video generation model, as open weights with both model weights and code under the Apache 2.0 licence. This means anyone can use, modify, and build on it commercially without licensing fees. It joins a growing wave of open-source video AI, following similar moves by other labs in recent months.

Why it matters: Open-weights video generation lowers the barrier for creators, educators, and developers to build video tools without depending on proprietary APIs. If you make content or teach using video, this matters to you.

Australia announces six mandatory AI safety guardrails for frontier developers

The Australian government announced six compulsory AI safety guardrails for companies developing frontier AI, with plans for a new regulator empowered to impose significant penalties. The move puts Australia alongside the EU and New York as jurisdictions with binding rules rather than voluntary guidelines.

Why it matters: More countries are moving from voluntary codes to enforceable law, and faster than most people expected. The global picture for AI regulation has shifted from "maybe someday" to "it is already happening in several places at once."

Japan is quietly debating a shift from AI guidelines to legal penalties

A closed-door government working group in Japan met on November 11 to discuss replacing the country's current voluntary AI guidelines with binding orders and legal penalties for AI developers. Japan has been one of the more permissive major economies on AI regulation. If it shifts, the pressure on companies operating globally gets meaningfully higher — and fast.

OpenAI launches India inference for GPT-6 Luna + Bihar and Goa announce state AI policies

OpenAI started in-country API inference for its GPT-6 Luna model in India for regulated enterprises, reducing latency and addressing data localisation concerns. Separately, Bihar and Goa both passed state-level AI policies this month — Bihar on November 2 and Goa's covering 2026–2029 — with dedicated funds and startup incentives. India's AI policy is now also being shaped at the state level, not just nationally.

All daily AI briefs