Gemini Omni goes live for developers, tech layoffs pass 100,000, and OpenAI vs. Anthropic just drew a sharp line over whether AI should run ads.
Gemini Omni is live: give it anything, it gives you video back
Google's Gemini Omni model went live today for developers via the Gemini API and Google AI Studio. Unlike earlier models, Omni takes any input (text, image, audio, or video), and can return video, image, or text outputs. It sits alongside Gemini 3.5 Flash, now generally available after its I/O debut. Google also opened a $2M Build with Gemini XPRIZE hackathon to developers globally.
Why it matters: True multimodal output (not just reading or describing media but generating it) is a qualitative shift in what everyday AI tools can do. Students, creators, and educators will likely feel this first as video generation stops being a specialist skill.
Tech layoffs pass 100,000 in 2026, nearly half linked directly to AI
Meta's 8,000-person reduction, announced May 20, is executing this week. Oracle eliminated 30,000 roles in the largest single cut of the year. Across the industry, 113,000+ tech workers have been laid off so far in 2026, with roughly 50,000 of those cuts explicitly attributed to AI-driven restructuring. Companies are shrinking traditional headcount to free capital for AI infrastructure: Meta alone is committing $115–135 billion this year.
Why it matters: This is not a slow transition. The job market is shifting faster than most upskilling programs are designed to handle. If you are in a tech-adjacent role and haven't actively built AI skills in the last six months, that gap is closing faster than it looks.
ChatGPT now runs ads. Anthropic says Claude never will.
OpenAI opened its ChatGPT Ads Manager beta to all businesses in the US, Canada, Australia, and New Zealand, removing the previous $50,000 minimum spend and switching to cost-per-click bidding. Simultaneously, Anthropic publicly reaffirmed that Claude will stay ad-free, framing it as a trust decision, not just a revenue one. Anthropic's ARR reportedly hit $30 billion in April, roughly 80% from enterprise and API customers, giving it a very different business model than OpenAI.
Why it matters: The AI you talk to every day is now likely to be showing you ads paid for by a third party, depending on which one you use. Two leading companies are making opposite bets on what AI users will tolerate. This distinction is worth knowing before you decide where to ask your next question.
30 US states now have active deepfake laws, and enforcement is live
As of this month, 30 US states have enacted legislation specifically targeting deepfakes in political communications and non-consensual intimate imagery, up from 28 at the start of 2026. Washington State and Pennsylvania's laws took effect this week. At the federal level, the TAKE IT DOWN Act requires platforms to remove non-consensual deepfakes within 48 hours of a valid notice. EU platforms face a separate deadline: machine-readable labels on AI-generated content become mandatory on August 2, 2026.
Why it matters: For the first time, deepfake harms carry legal consequences, not just policy flags. For students and working professionals, this changes what's legally actionable if you are a victim, and what's risky if you are creating synthetic media carelessly.
IEA: AI will triple electricity demand by 2030. Nobody has a full plan.
The International Energy Agency projects AI-related data centre electricity consumption will triple by 2030. US data centres are on track to jump from 80 to 150 gigawatts of power demand by 2028, the equivalent of adding Spain's entire electricity footprint in three years. Meta's $115–135 billion AI infrastructure commitment this year is part of a spending race across big tech that is straining grid capacity in the US, Europe, and South Asia.
Why it matters: Energy is fast becoming the hard ceiling on how quickly AI can grow. For India (which is building 38,000+ GPUs into its national AI infrastructure) this is a live policy challenge. AI ambitions and power supply have to scale together or one caps the other.
India is using AI to fix its water crisis, and it starts in Bengaluru today
India's first National AI and Digital Water Summit opened today in Bengaluru, hosted by the Bangalore Water Supply and Sewerage Board with 400+ stakeholders from government, technology, and urban planning. The headline launch: an AI-powered command centre for real-time water and sewage management, covering leak detection, energy optimisation, and data governance. Inaugurated by Chief Minister Siddaramaiah. This is what "AI for public good" looks like when it moves off a conference slide and into infrastructure.
AI study tools raise $38M in a month + EU deepfake labels arrive August 2
Gizmo, an AI-powered study app with 13 million users, closed a $22M Series A. Oboe, which builds personalised AI course curricula in seconds, raised $16M. Both signal investors are rotating back into consumer edtech, but only for AI-native tools with real traction. Separately, the EU's AI Act requirement for machine-readable labels on AI-generated content takes effect August 2, 2026. Platforms generating images, video, or text at scale have roughly two months to comply.