Google Overhauls DeepMind as OpenAI, Anthropic Surge
ALSO: AI now decides who’s laid off, ChatGPT reads labs
Welcome to Guru99 AI Report!
Top Story: Google just shook up DeepMind, AI is quietly deciding who gets laid off, and ChatGPT now reads your medical records. This week’s shifts are big — and closer to your world than you’d think. Let’s dive in.
🔄 Google Overhauls DeepMind as Rivals Pull Ahead
Brief Buzz:
Google just reset the top of its AI unit. On Wednesday, Alphabet named a new operational chief for Google DeepMind, moved co-founder Demis Hassabis into a big-picture role, and lost longtime engineering leader Jeff Dean to a startup — all amid intense pressure from OpenAI and Anthropic.
- Demis Hassabis shifts from CEO to chairman of Google DeepMind and Alphabet chief scientist, focusing on AGI research.
- Koray Kavukcuoglu, DeepMind’s CTO, takes over day-to-day operations as SVP of Google DeepMind, reporting directly to CEO Sundar Pichai.
- Jeff Dean exits Google after 27 years to co-found Discovery Loop, an AI-for-science startup backed by Alphabet.
- The reshuffle follows repeated delays to flagship Gemini 3.5 Pro and a wave of researchers leaving for OpenAI and Anthropic.
- Google frames the moves as accelerating its AI work, not a retreat, and says Dean’s exit is unrelated.
💡 Why Should You Care?
Google still owns huge advantages — compute, cloud, and Android and Workspace’s billions of users. But leadership churn plus shipping delays suggest it’s playing catch-up.
🚨 AI Agents Are Flooding Security Teams
Brief Buzz:
Cyberattacks are moving faster, and AI agents are adding to the pressure. According to CrowdStrike’s 2026 Threat Hunting Report, security teams are seeing more agent-triggered leads, while some flaws are exploited within hours. Cloud identities and AI supply chains are also under attack. The old security playbook is starting to break.
- 2.5x: AI agent-triggered detection leads tracked at 2.5 times the rate of human-triggered leads — not confirmed attacks.
- 88%: In January–June 2026, CrowdStrike-observed exploitation of vulnerabilities with public proof-of-concept code occurred within 48 hours of release.
- 171%: Cloud-conscious eCrime — CrowdStrike’s term for financially motivated cybercrime — surged over the report’s 12-month observation period.
- Separate supply-chain campaigns poisoned at least 131 Mastra AI framework packages and compromised 300+ dependencies in one day.
💡 Why Should You Care?
For businesses, that means patching the most exploitable flaws first and tightening controls around AI access, cloud identities, API keys, and developer pipelines.
🤖 AI Is Entering the Layoff Room
Brief Buzz:
AI is no longer just drafting emails. It is entering the room where jobs are cut. A July 2026 vendor survey of 1,000 U.S. managers already using AI for direct-report decisions found that many consult it on layoffs. Some occasionally let it decide without human review.
- 59% said AI helps decide who gets laid off; 24% use it often or all the time.
- 43% sometimes allow unsupervised AI layoff decisions, though 91% would override a recommendation they disagreed with.
- Among AI-for-layoff users, inputs included performance (80%), attendance (57%), medical leave or sick days (31%), tenure (32%), and age (14%).
- 38% had no ethical-AI training; 58% could not confirm whether their company tested the system for bias.
💡 Why Should You Care?
U.S. anti-discrimination laws still apply to AI-assisted employment decisions, while protected-leave rules bar retaliation. That makes documented criteria and accountable human review essential.
🎬 Google’s Flow Films Find AI’s Creative Sweet Spot
Brief Buzz:
Google’s fourth Flow Sessions cohort makes a stronger case for AI filmmaking than fully synthetic spectacle. Its most persuasive example combines live production with targeted generation, treating AI like a flexible visual-effects tool rather than a replacement for actors, sets, or human direction.
- Six weeks: Google’s creative partner program gave selected artists time to build passion projects with its Flow studio.
- Hybrid, not all-AI: Filmmaker Andrew Nethery used live action and AI together for Nothin’ Beats a Cheeseburger. LinkedIn
- A deliberate fit: The trailer follows an alien seeking Earth’s best burger — fantasy imagery without a real-world benchmark.
- Under three weeks: Nethery says his team shot and delivered the film that quickly, a first-party production claim. LinkedIn
💡 Why Should You Care?
Nethery’s reported timeline suggests AI may shorten some VFX workflows while preserving human performances. Copyright and licensing and disclosure questions still require scrutiny.
🩺 ChatGPT Now Plugs Into Your Medical Records
Brief Buzz:
OpenAI has taken Health in ChatGPT live for U.S. adults, letting them connect Apple Health, hospital portals, One Medical, or Function Health. The bigger shift: health context now follows you into any chat, not just a walled-off Health tab.
- Live now for logged-in U.S. users 18+ on web and iOS, across Free, Go, Plus, and Pro plans.
- Over 70% of health chats happened outside the earlier dedicated Health space, so OpenAI removed the wall.
- Connected records and chats using them are excluded from model training and ad targeting, OpenAI says.
- 300 million people ask ChatGPT health questions weekly, per OpenAI’s own figure.
- A Florida pastor sued OpenAI over allegedly dangerous health advice one day before launch.
💡 Why Should You Care?
A polished answer built on your real labs feels more authoritative than a generic one. Use it to prep questions for a clinician, not to replace one.
💸 OpenAI Bets $750 Billion on Owning Its Compute
Brief Buzz:
OpenAI now expects to spend $750 billion on compute through 2030 — roughly 25% above its earlier guidance — and is building its first data center itself. Rival Anthropic is scaling just as hard, but mostly by renting other companies’ chips.
- $20 billion kicks off Project Camellia, a 3.2-gigawatt Georgia campus OpenAI is designing and building itself.
- $45 billion buys Anthropic three years of SpaceX’s Colossus clusters; a $10 billion Meta deal is still just talks.
- AMD will ship 2 gigawatts of GPUs to Anthropic and invest up to $5 billion back into it.
- Capacity is tight: Anthropic caps flagship Fable 5 at half of weekly usage limits, even on Max.
- Secured compute is now a moat, argues Apollo partner Rob Bittencourt in a June research note.
💡 Why Should You Care?
Compute scarcity now shapes what you can actually use: tighter usage limits, higher prices for frontier models, and gigawatt-scale data centers landing in communities already arguing over power and water.
Hey! I’m Krishna Rungta
Founder of Guru99.com, Editor-in-chief & Technology Expert
Was this email forwarded to you? Sign up for free here.
