This week in AI - June [18]

Check out our weekly AI news discussing OpenAI striking a shocking cloud deal with Google despite their fierce rivalry, Apple focusing on practical AI over grand promises at WWDC, and Meta's AI app becoming a privacy disaster with users accidentally sharing intimate details publicly. You'll also find that Amazon's CEO warned of corporate job cuts while Nvidia's chief disputed doom-and-gloom automation predictions.

🔔 HEADLINE MAKERS

Apple WWDC 2025 [Keynote][Announcement][TheNewYorkTimes][CNBC]

• Refined Focus: Apple shifted gears at WWDC 2025, toning down last year's ambitious AI rhetoric to focus on practical, everyday features. 

• Key Capabilities: Visual Intelligence now works on iPhone screen content, allowing users to search Google or ask ChatGPT about anything they're viewing in apps [TechCrunch]. Live Translation powers real-time conversations across Messages, FaceTime, and phone calls, running entirely on-device [TechCrunch]. Workout Buddy brings AI coaching to Apple Watch using voice data from Fitness+ trainers to deliver personalized motivation during exercise sessions [TechCrunch].

• Developer Access: The Foundation Models framework opens Apple's on-device AI to third-party developers with just three lines of Swift code, enabling offline AI features without cloud costs. [Apple research]

• Siri Setback: The most disappointing news was that promised AI-powered Siri improvements won't arrive until next year, with Apple's Craig Federighi offering no timeline for the enhanced voice assistant capabilities that attendees were eager to see. Apple Execs Defend Siri Delays [Interview by WSJ][TechCrunch]

OpenAI Partners with Google Cloud Despite AI Rivalry [Reuters]

• Unprecedented Deal: OpenAI has finalized a cloud computing partnership with Google, marking a surprising collaboration between two fierce AI competitors. The deal, completed in May after months of negotiations, allows Google Cloud to supply additional computing capacity for training and running OpenAI's AI models alongside their existing Microsoft infrastructure.

• Strategic Diversification: The partnership represents OpenAI's continued effort to reduce dependency on Microsoft Azure, which had been their exclusive cloud provider until January. With ChatGPT's annualized revenue hitting $10 billion as of June, OpenAI is aggressively expanding compute capacity through deals with Oracle, SoftBank's Stargate program, and CoreWeave, while also developing their first in-house chip.

• Market Implications: Google's stock jumped 2.1% on the news while Microsoft dropped 0.6%, with analysts calling it a "big win" for Google Cloud despite ChatGPT threatening Google's search dominance. The deal highlights how massive AI computing demands are reshaping competitive dynamics, forcing rivals to collaborate even as they compete directly in consumer AI applications.

• Capacity Tensions: The partnership creates complex resource allocation challenges for Google CEO Sundar Pichai, who must balance selling compute power to rivals against Google's own AI development needs. Google Cloud already faces capacity constraints while trying to demonstrate returns on $75 billion in expected AI capital expenditures this year.

AI-Generated Advertising Push Accelerates

• Amazon Video Generator: Amazon has launched its enhanced AI video generator to all U.S. advertisers after nine months in beta, evolving from basic clips to sophisticated high-motion videos showing products in realistic action. The tool creates multi-scene videos with six options per request, transforms still images to motion with one click, and includes video summarization features - though critics argue flooding the marketplace with AI-generated ads contradicts quality commitments and raises authenticity concerns on a platform already struggling with fake reviews. [Announcement][AndroidPolice]

• TikTok is expanding its AI ad tool Symphony with text-to-video generators that create influencer-style content without actual influencers, featuring digital avatars that can model clothes, hold products, and demonstrate apps. TikTok is going to be filled with video ads generated with AI, some featuring individuals who have never existed. The platform aims to blur the line between paid and organic content by helping brands match TikTok's native visual style, raising questions about the future of creator economics as AI avatars require no contracts, fees, or negotiations while producing endless content streams.[Announcement][Adweek]

• Reddit Community Intelligence: Reddit has launched two new ad tools powered by insights from over 22 billion posts and comments - Reddit Insights for social listening and conversation summaries, and Conversation Summary Add-on that lets brands showcase positive user sentiment below their ads. The company is betting on community marketing and AI-powered insights as it reported $358.6 million in Q1 ad revenue, up 61% year-over-year, positioning itself as a credible source of human conversation amid the flood of AI-generated content online. [Announcement][Axios]

🌪️ AI IN THE WILD

TV Milestone: $2,000 AI Ad Airs During NBA Finals [The Verge][Kalshi's tweet] 

• Betting platform Kalshi aired a fully AI-generated advertisement during the NBA Finals that cost just $2,000 to produce using Google's Veo 3 text-to-video generator, featuring surreal scenes of elderly cowboys with chihuahuas, people swimming in pools of eggs, and aliens drinking beer. The ad required 300-400 generations to produce 15 usable clips over 2-3 days by one AI filmmaker, representing a claimed 95% cost reduction compared to traditional advertising production methods.

Meta AI App Privacy Disaster [TechCrunch][BusinessInsider]

• Meta's standalone AI app has become a privacy nightmare as users unknowingly share intimate conversations publicly, including tax evasion questions, family legal troubles with full names, home addresses, and medical concerns. The app's "Discover" feed reveals deeply personal interactions like grief counseling, child custody battles, birthday poems for wives, and even pocket-dial recordings where Meta AI interrupts real conversations, with users' real Instagram handles attached to posts they may not have intended to share publicly despite the multi-step sharing process.

Microsoft 365 Copilot Zero-Click Vulnerability [Aim Labs]

• Critical Security Flaw: Security researchers have discovered "EchoLeak," a zero-click AI vulnerability in Microsoft 365 Copilot that allows attackers to automatically exfiltrate sensitive organizational data without user interaction by simply sending a malicious email. The attack exploits "LLM Scope Violation" - where untrusted external emails containing disguised instructions make the AI access privileged internal data including mailboxes, OneDrive files, SharePoint sites, and Teams chats, then automatically transmit the most sensitive information to attacker-controlled servers through bypassed security controls including link redaction and content security policies.

Google DeepMind Revolutionizes Hurricane Forecasting [VentureBeat]

• Google DeepMind has partnered with the U.S. National Hurricane Center to deploy an AI system that predicts hurricane paths and intensity with unprecedented accuracy. The model generates 15-day forecasts in one minute versus hours for traditional systems, with predictions 140 kilometers closer to actual storm positions than existing models - a breakthrough that could save lives during the 2025 hurricane season.

🤖 AI VS. HUMAN WORKERS

Nvidia’s Jensen Huang says he disagrees with almost everything Anthropic CEO Dario Amodei says [Fortune]

• Nvidia CEO Jensen Huang isn’t sure about Anthropic CEO Dario Amodei’s recent predictions about AI-driven job automation. Speaking at VivaTech in Paris, Huang pushed back on the idea that AI could soon replace half of all entry-level office roles and questioned the philosophy behind limiting AI development to a few actors.

AI-Driven Workforce Reduction [Amazon]

• Amazon CEO Andy Jassy announced that extensive AI adoption will reduce the company's corporate workforce in coming years as the company builds over 1,000 generative AI services and applications. Jassy emphasized that while some jobs will become obsolete, AI agents will change work scope and speed, allowing teams to focus less on routine tasks and more on strategic innovation, positioning Amazon to "operate like the world's largest start-up" with leaner, more efficient teams.