Daily AI Brief: November 14, 2026

India's AI rules enter public consultation, NVIDIA's leadership reportedly changes hands, and the EU's content-marking deadline arrives in 18 days.

Jensen Huang reportedly steps down as NVIDIA CEO after 33 years

An SEC Form 8-K filing indicates Jensen Huang has resigned as NVIDIA's chief executive. CFO Colette Kress is named interim CEO, and Harvey Jones is listed as the board's first chair. NVIDIA is due to report Q3 fiscal 2027 results on November 17. The news has not been confirmed by NVIDIA's official communications channel; a spokesperson said as recently as November 9 that Huang remains committed to leading the company without interruption.

Why it matters: Jensen Huang built NVIDIA into the most important company in the AI infrastructure stack. If confirmed, this transition reshapes every conversation about chips, data centres, and the pace of AI hardware development. Watch for official statements in the coming hours.

India's first AI safety consultation is now open — and it covers deepfakes

MeitY released its long-awaited AI consultation paper this week, opening a public comment window on AI safety, deepfake accountability, and startup compliance. The government is also reportedly pushing central departments in defence and aerospace toward indigenous AI models and away from foreign ones. Bihar's state cabinet separately approved the Bihar AI Policy 2026, which targets AI investment and skilled jobs. TRAI has also proposed AI accountability rules for telecom operators.

Why it matters: India is not waiting for a single national law — it is building policy in layers, through the central consultation, state-level plans, and sector regulators all at once. If you are building or studying AI in India, the consultation window is the place to make your voice heard.

EU content-marking deadline: 18 days. Platforms scramble.

From December 2, 2026, the grace period for content labelling under the EU AI Act ends for providers of generative AI systems already on the market. Any AI-generated image, audio, or video must carry a machine-readable marker. Platforms that have been delaying compliance have roughly two weeks to get their pipelines in order.

Why it matters: This is not a "coming soon" rule — the enforcement clock is running now. Creators, brands, and publishers operating in Europe need to verify their content workflows today, not in January.

New US Congress plans AI hearings. New York's RAISE Act is live.

Democrats won both chambers in the November 4 US midterms and are signalling quick action on AI. Plans include new hearings with leading AI labs and proposals for what they're calling "reasonable guardrails." Separately, New York's RAISE Act came into effect this month: it requires frontier AI developers operating in the state to register and report safety incidents within 72 hours beginning in January 2027.

Why it matters: The US regulatory picture is shifting. After years of a relatively light federal touch, a new Congress with an AI agenda and a state-level law with actual teeth means 2027 will look different for every company building at the frontier.

One in three educators doesn't know their school's deepfake policy

A new Education Week Research Center survey found that 33% of educators don't know whether their school has a policy addressing harmful AI-generated images. 13% said they don't know what a deepfake is. Fake student and teacher images are spreading across school social media groups. Google separately announced it will update Search access for schools after backlash over a plan that would have cut students' search access in AI-ban districts.

Why it matters: Deepfakes are already in classrooms. The gap is not access to tools — it is the absence of basic training for teachers and clear policy for students. This is becoming a school safety issue, not just a tech issue.

Google's Gemini 4 Hydrogen is here — built for speed and coding agents

Google released Gemini 4 Hydrogen this week, a fast model optimised for coding tasks and agent workflows. It sits alongside Gemini Decision 1, a model launched earlier this week for fast, closed-set choices such as routing and classification. IBM also released Granite 4.1 LLMON-preview models in 3B and 8B sizes, targeting enterprise deployment.

Why it matters: Specialised, faster models are becoming the new competition battleground — not just the largest ones. For developers building AI-powered tools, the choice of which model to use for which task is now more nuanced and more important.

The US-China chip war just sharpened: no H100-equivalent for China

US Treasury Secretary Scott Bessent stated this week that China should not receive chips more advanced than NVIDIA's H100. Export controls are an active legislative fight in Washington, and with a new Congress arriving and the NVIDIA leadership situation unresolved, every decision about who gets access to AI compute is under fresh scrutiny. The H100 itself is already one generation behind what US labs are running. The gap this draws will shape what China can build — and how fast.

Wabi raises $55M at $1.25B; Japan moves toward binding AI safety orders

Consumer AI assistant startup Wabi announced a $55 million seed round led by IVP at a $1.25 billion valuation — one of the largest seed rounds in AI this year. Separately, Japan's government working group met on November 11 to discuss binding orders and legal penalties for advanced AI developers that ignore safety rules, a significant shift from its previous guidelines-only approach.

All daily AI briefs