Google search merged with AI Mode for 1 billion users, OpenAI's health AI cuts errors by 71%, and deepfake job fraud is now a documented crisis.
Google Search is now an AI interface, and 1 billion people just switched
Google officially unified AI Overviews and AI Mode into a single search experience, now powered by Gemini 3.5 Flash. The redesigned Search handles conversational queries, live image and video questions, and AI agents that can track information and complete tasks on your behalf. One billion people use AI Mode monthly, with query volume doubling every quarter since launch.
Why it matters: This is not a new feature toggle, it's the end of the blue-link search page most people have used for 25 years. If you've been treating AI Mode as optional, Google just made it the default for everyone.
OpenAI's health AI cuts factual errors by 71%, and it's free for everyone now
OpenAI updated ChatGPT's healthcare experience with GPT-5.5 Instant, reporting a 71% drop in factual errors on health-related queries after doctor-led testing. The upgrade is available to all free ChatGPT users. The older GPT-5.2 model has been fully retired from ChatGPT as of June 12, all existing conversations automatically continue on GPT-5.5.
Why it matters: A more accurate health AI at zero cost matters for anyone who has used ChatGPT to understand a diagnosis, a medication, or a test result. It is not a replacement for a doctor, but it is getting meaningfully more reliable for background research.
Anthropic is in talks to run Claude on Microsoft's own AI chips
Anthropic is in early-stage negotiations to run Claude inference workloads on Microsoft's custom Maia 200 AI accelerators via Azure. If the deal closes, Claude would become the first external frontier model to run on Microsoft's custom silicon, hardware that has until now only run OpenAI's own GPT models. Anthropic already uses four chip vendors and is building a fifth option.
Why it matters: AI performance is increasingly shaped not just by who writes the smartest model, but who controls the hardware it runs on. When the biggest AI labs start evaluating their own chip suppliers, it reshapes the competitive landscape for everyone below them.
1 in 3 job applications may be AI-generated fakes. It's already a crisis.
New research shows that more than one-third of analyzed job applicant profiles may be entirely AI-fabricated, synthetic resumes, AI-generated portfolio work, and deepfake video interviews capable of passing real-time screening. Gartner projects that 1 in 4 candidate profiles globally will be fake by 2028. Companies are now adding identity verification layers and in-person interview requirements in response.
Why it matters: If you're hiring, your pipeline may already be encountering this. If you're job hunting legitimately, stricter verification is coming regardless, the same AI tools behind the fraud are widely available, so employers are adjusting how they screen everyone.
IndiaAI opens its global startup accelerator: Cohort 2 applications are live
The IndiaAI Startups Global International Acceleration Program has opened its second cohort, offering selected Indian AI startups access to international acceleration programs and up to INR 5 lakh in prize money from the IndiaAI Mission. The program is part of India's broader ₹10,371 crore ($1.25B) AI investment strategy. Indian startups have attracted $8.44B in equity funding in 2026 so far, across 831 rounds.
Why it matters: Government-backed acceleration with international reach is an open door right now, not a press release. If you are building an AI startup in India and want global exposure, this cohort is a concrete opportunity with a deadline.
AI's electricity bill is growing faster than the clean grid can keep up
A new UN report puts hard numbers on something the industry has talked around: AI data centers consumed 448 terawatt-hours of electricity last year (more than all but 10 countries), and produced 208 million tons of CO₂ in the process. That could triple to 945 TWh by 2030. About 40% of the additional demand will still be met by gas and coal. Ireland has already paused new data center approvals near Dublin until 2028 because the grid cannot keep up. Watch for energy constraints to become the next real choke point in AI development, not model capability, not compute, but whether the lights stay on.
Meta's paid social subscriptions go global + NAVER and NVIDIA sign a gigawatt-scale AI deal
Meta officially launched consumer subscription plans worldwide: Instagram Plus ($3.99/mo), Facebook Plus ($3.99/mo), and WhatsApp Plus ($2.99/mo): signalling that ad-only revenue is no longer the whole plan. Separately, South Korea's NAVER and NVIDIA announced a sovereign AI infrastructure partnership starting at 55 megawatts and scaling to gigawatt capacity at NAVER's Sejong data center, with a new AI Agent Platform launching in Korea in the second half of 2026.