Weekly AI News

Hey it’s Jul,
Greetings and welcome to the twelfth edition of “Weekly AI News”!

First, OpenAI reversed its own reversal on the o3 model (yes, it’s a bit funny to say) while delaying GPT-5 to secure extra server capacity and handle the massive demand they’re anticipating.

Meanwhile, Google is pushing full steam ahead: with its Ironwood chip, chatbots are getting even snappier, and the new Agent2Agent protocol acts like a universal “language” to connect every AI agent out there. In other words, nobody’s satisfied with merely launching a single model anymore—everyone wants to be the platform you simply can’t live without.

And while all that brainstorming is going full tilt in California, Europe is stepping up with a muscular action plan to catch the frontrunners, keen on building its own AI ecosystem rather than relying solely on American tech giants.

The result isn’t just a contest of models but rather a battle for who sets the rules. The message is clear: who can master the best infrastructure and become the standard everyone flocks to? Stay tuned—coffee in hand and critical thinking on standby!.

Happy reading!

💡🚀 OpenAI to Release o3 After All, Delays GPT-5

OpenAI plans to launch both o3 and o4-mini in the coming weeks, reversing its previous cancellation of the o3 model. The rollout of GPT-5 is postponed to ensure enough capacity to meet unprecedented demand. GPT-5 will incorporate voice, Canvas, search, deep research, and more.

🧠✨ OpenAI Gives ChatGPT a Memory Boost

OpenAI just unveiled a major update that allows ChatGPT to automatically recall user preferences across all conversations. The assistant continuously “listens” to user context and adjusts responses for more relevant answers. Users can easily opt out or modify what ChatGPT remembers about them, offering a balance between personalization and privacy.

🖼️🔏 OpenAI Tests Watermarking for ChatGPT-4o ImageGen

OpenAI is exploring a watermark for images generated by ChatGPT-4o’s ImageGen feature, especially on free accounts. Paid subscribers might save watermark-free images, though the plan could change. The move aims to address concerns over copyright and authenticity, given the growing popularity of generating artistic styles like Studio Ghibli.

💼🤝 OpenAI Considers Buying Hardware Startup from Jony Ive and Sam Altman

OpenAI has held talks to acquire io Products, a new AI-powered device venture involving Sam Altman and ex-Apple designer Jony Ive. A proposed deal could be valued at over $500 million, bringing in a team working on screenless phones and AI-driven household products. This acquisition might accelerate OpenAI’s consumer hardware efforts and pit it more directly against Apple.

🔐💸 OpenAI Invests in Adaptive Security

In its first foray into cybersecurity, OpenAI has invested in Adaptive Security, a startup that uses AI to simulate and train employees against social engineering attacks. The company recently secured $43 million in Series A funding and already counts more than 100 clients since launching in 2023.

🇪🇺💶 OpenAI Publishes ‘EU Economic Blueprint’

OpenAI released its vision for Europe, proposing a €1B AI accelerator fund and seeking to train 100 million Europeans in AI skills by 2030. The plan highlights innovation, economic growth, and responsible AI development. It aims to strengthen the EU’s position as a global technology leader.

⚖️⌛ Musk v. OpenAI Trial Delayed to March 2026

Elon Musk’s lawsuit against OpenAI, challenging its transition to a for-profit entity, has been postponed to March 2026. OpenAI is currently raising $40 billion, with $20 billion contingent on converting to a for-profit. If OpenAI proceeds and loses the trial, it may be forced to revert to nonprofit status.

⚔️🚫 OpenAI Countersues Musk

OpenAI filed a countersuit accusing Elon Musk of “malicious campaigns” and “unfair competition,” including a “sham” $97 billion takeover bid. The company seeks damages and an order barring further interference. Musk’s initial suit argued OpenAI violated charitable duties by shifting to a for-profit structure, while also questioning SoftBank’s conditional investment in OpenAI.

💲📡 xAI Launches Grok-3 API

Elon Musk’s xAI is making its flagship Grok-3 model available via API at $3/$15 per million tokens, plus a lower-cost Grok 3 Mini option. Grok-3 handles images, Q&A, and powers features on X. Faster tiers cost more, reflecting xAI’s ambition to challenge Google and OpenAI in the multimodal AI space.

☁️⚙️ Big AI Day at Google Cloud Next 2025

Google unveiled a slew of AI updates: an agent-based coding platform, advanced AI chips (“Ironwood”), next-level image and audio models, and a faster, cheaper Gemini 2.5 Flash. Workspace enhancements include AI for Docs (audio summaries), Sheets (data analysis), and automated flows bridging Forms, Chat, Gemini, and more. Google demonstrates its commitment to AI across every product line.

⚙️💡 Google Unveils Its Ironwood Inference-Focused Chip

Google introduced Ironwood, its seventh-generation Tensor Processing Unit, optimized purely for inference tasks (e.g., chatbot queries). Unlike previous generations aimed at both training and inference, Ironwood targets faster, more efficient responses. It strengthens Google’s ongoing strategy of developing in-house AI accelerators to rival Nvidia’s dominant GPU offerings.

🤖🌐 Google Introduces Agent2Agent (A2A) Protocol

Google launched A2A, an open protocol that enables AI agents from various platforms to communicate and cooperate. Supported by over 50 major companies like Salesforce and PayPal, A2A helps agents discover each other’s skills and collaborate without sharing memory. It complements Anthropic’s MCP for tool integration, aiming to unlock complex multi-agent workflows.

📱🎥 Google Expands Gemini Live Video Features

Google is expanding Project Astra’s real-time visual AI to more Android devices, including Pixel 9 and Samsung Galaxy S25. Users can have multilingual conversations with Gemini about anything captured on camera or through screen sharing. Early tests show partial real-time analysis rather than continuous video, but it’s another step toward more immersive AI-driven experiences.

🔎🔗 Google Extends AI Mode and Krea Raises $83M

Google extended AI Mode to a broader user base, adding multimodal search to interpret both text and images via Gemini and Google Lens. Meanwhile, the creative platform Krea secured $83 million in funding to add audio and enterprise features. Both announcements highlight the ongoing innovation frenzy within the AI ecosystem.

💰🛑 Google Allegedly Pays Some AI Staff to Do Nothing

To retain top AI researchers amid fierce competition, Google’s DeepMind reportedly enforces strict noncompete clauses in the UK. Some employees are paid for up to a year to prevent them from joining rivals like OpenAI. While this keeps talent away from competitors, it also sidelines experts from rapid AI advances.

🦙⚡ Meta Introduces Llama 4 with Two Models Available

Meta released Llama 4, offering two new models—Scout and Maverick—while teasing two more: Behemoth and Reasoning. Scout (109B parameters) can run on a single H100 GPU, and Maverick (400B) has a 1M token context window, outperforming GPT-4o on key benchmarks. Both use mixture-of-experts architecture to reduce computation costs and are available now.

❌🎯 Meta Exec Denies Alleged Llama 4 Score Inflation

Rumors on X and Reddit claimed Meta boosted Llama 4’s benchmark scores artificially. Meta’s VP of Generative AI, Ahman Al-Dahle, denied these claims, saying the models weren’t trained on test sets. The gossip appears to stem from an anonymous post by someone claiming to have resigned from Meta over its benchmarking practices.

⚙️📝 Copilot Gets Personalization Upgrades

Microsoft Copilot now remembers user details, routines, and contexts across conversations, offering more tailored responses. New “Actions” let Copilot handle web tasks like restaurant bookings and ticket purchases, while Copilot Vision integrates real-time camera input. The aim is a deeper, more proactive AI assistant that blends work and personal life.

🔝💼 Anthropic Unveils “Claude Max” Plan

Anthropic introduced a premium tier for Claude AI—Claude Max—offered at $100/month for 5× the Pro limits or $200/month for 20×. Subscribers get priority access to new features and models. Despite the higher price, usage caps remain, prompting mixed reactions from users who want unlimited usage.

🎨🖥️ Adobe Builds AI Agents for Photoshop and Premiere Pro

Adobe revealed it’s developing “creative agents” to suggest context-aware edits in Photoshop and Premiere Pro, then perform them automatically. A new floating Actions panel will recommend tasks like removing background objects or enhancing focus. This complements existing AI tools such as generative fill and advanced object removal.

🖌️📈 Canva Debuts Visual Suite 2.0 and AI Design Partner

At Canva Create 2025, the company launched Visual Suite 2.0—a single platform for documents, websites, and more. Canva Sheets turns spreadsheets into interactive stories, while Magic Charts and Magic Studio offer bulk content generation. Canva Code allows users to create interactive designs from prompts, all aiming to speed up collaboration at scale.

🎬⚡ Runway Releases Gen-4 Turbo

Runway unveiled Gen-4 Turbo, a faster version of its AI video model capable of generating 10-second clips in just 30 seconds. This speed boost could benefit content creators needing quick video prototypes and real-time marketing assets. The release underscores Runway’s continued focus on professional-grade AI video editing.

🔊🌐 ElevenLabs Integrates MCP for AI Voices

ElevenLabs, known for its realistic voice synthesis, now supports the MCP protocol, enabling external AI systems like Claude to tap into its voice-generation services. Developers can build automated agents that combine Claude’s text generation with ElevenLabs’ speech. This opens the door to more advanced voice-based assistance and customer service tools.

👻📸 Snapchat Introduces Sponsored AI Lenses

Snapchat rolled out AI-driven Lenses for advertisers, allowing brands to create interactive AR experiences that adapt in real time to user features. By blending sponsored content with AI personalization, these Lenses turn users into playful brand ambassadors. The feature expands the platform’s already robust AR marketing capabilities.

🔈🎥 Amazon Announces Nova Sonic and Upgrades Nova Reels

Amazon launched Nova Sonic, a new voice model delivering natural speech with just 1.09 seconds of latency, plus Nova Reels 1.1, supporting video generation up to two minutes. Both models are available through Amazon Bedrock at lower prices than competing solutions. It highlights Amazon’s push to stay competitive in voice and video AI.

📽️🏆 Nvidia & Stanford Achieve One-Minute AI Cartoons

Researchers from Nvidia and Stanford introduced “Test-Time Training” (TTT), a technique that lets AI generate minute-long cartoon animations with character and scene consistency. By adding TTT layers as a “memory,” the system outperforms traditional models. It marks a leap toward longer-form, coherent storytelling in AI-generated video.

💻💸 Nvidia Finalizes Acquisition of Lepton

Nvidia has completed its purchase of Lepton AI, a two-year-old startup offering rentable GPU servers for AI workloads. The deal is reportedly in the hundreds of millions of dollars. Lepton’s founders, both ex-Alibaba and Meta, join Nvidia to bolster its strategy of expanding GPU-based cloud services alongside direct chip sales.

🔬🤏 ‘World’s Most Advanced Microchip’ Unveiled

TSMC introduced a 2nm chip, set for mass production later this year. It promises a 10–15% speed boost at the same power or 20–30% reduced power at the same speed compared to 3nm designs, with a 15% increase in transistor density. This leap could fuel next-gen devices with enhanced performance and efficiency.

🚖🛡️ Waymo Explores Using Rider Data for Generative AI

Researcher Jane Manchun Wong found an unreleased Waymo One privacy page suggesting interior camera data could train AI or personalize ads. Waymo confirms users will be able to opt out but denies any intent to target ads. The move raises questions about privacy in autonomous vehicles and data use for AI modeling.

👓🤖 ByteDance Develops AI-Powered Smart Glasses

ByteDance, parent company of TikTok, is reportedly building AI-driven smart glasses to rival Meta’s Ray-Ban Stories. The device could capture photos, videos, and enable AI chatbot interactions. It underscores growing interest among global tech giants in hardware that integrates advanced AI features for immersive experiences.

⚽🏠 Samsung’s Gemini-Powered Ballie Home Robot

Samsung and Google are joining forces to launch Ballie, a small rolling home robot equipped with Google’s Gemini AI. Ballie can project videos, manage connected devices, and respond to voice commands. First revealed at CES 2020, the device is set to debut in the U.S. and South Korea this summer, possibly pioneering a new wave of consumer robotics.

🤝🔬 Murati’s Thinking Machines Adds Ex-OpenAI Talent

Mira Murati’s AI startup, Thinking Machines Lab, welcomed ex-OpenAI CRO Bob McGrew and GPT architect Alec Radford as advisors. Half of the startup’s founding team are OpenAI veterans, including co-founder John Schulman. While the firm remains secretive about its products, the addition of key AI pioneers raises anticipation for a major launch.

💰🏗️ Mira Murati’s AI Startup Eyes Record Funding

Thinking Machines Lab, led by ex-OpenAI CTO Mira Murati, is reportedly seeking $2B at a valuation of at least $10B—doubling its initial goal. The team’s OpenAI pedigree appears to be fueling investor excitement, despite no public product or revenue plan. Meanwhile, Ilya Sutskever’s SSI is also pursuing a lofty $30B valuation, reflecting the ongoing AI funding arms race.

🐞🔧 Study: AI Still Struggles with Debugging

A Microsoft Research paper shows that top coding models, including Anthropic’s Claude 3.7 Sonnet, struggle with real-world debugging tasks. Tested on 300 issues from SWE-bench Lite, the best model solved less than half. Researchers suspect the lack of human “debugging trace” data in training sets is holding back AI’s problem-solving prowess.

🏦🤖 Shopify CEO Demands Teams Prove AI Can’t Do Their Jobs

Shopify’s Tobi Lütke told staff they must demonstrate why AI can’t perform certain tasks before adding headcount or budget. Employees are also evaluated on their AI proficiency, with Lütke calling AI usage a “fundamental expectation.” The move aims to boost efficiency while ensuring humans stay indispensable.

⚠️🌍 ‘AI 2027’ Warns of ASI Risks

Ex-OpenAI researcher Daniel Kokotajlo and the AI Futures Project predict human-level AI within two years, leading to superintelligence by 2027. The report outlines scenarios where accelerated progress triggers an intelligence explosion, posing existential risks. It underscores urgent calls for robust governance and safety measures before AI outpaces human control.

🏢💲 Andreessen Horowitz Plans $20B AI Fund

Venture capital giant a16z reportedly aims to raise $20 billion—its biggest fund to date—to invest in mature AI startups. The firm has already backed projects like coding assistant Cursor and Elon Musk’s xAI. The new round would significantly surpass its previous $7.2B raise, reflecting the escalating appetite for AI investments.

⚖️🕵️ Democratic Senators Question Microsoft, Google Deals

Senators Elizabeth Warren and Ron Wyden have queried whether multibillion-dollar investments in OpenAI and Anthropic violate antitrust laws. They probe how these deals might give Microsoft and Google undue control or funnel startup spending to their own cloud services. The inquiry mirrors Elon Musk’s accusations of conflicts of interest at OpenAI.

🇪🇺🏗️ EU Launches €200B “AI Continent Action Plan”

The European Commission announced a massive initiative to build 13 AI “factories” and triple data center capacity in the EU within seven years. It also seeks to streamline regulations for startups hampered by heavy compliance costs. This strategy aims to bolster Europe’s global AI standing while balancing innovation with public safeguards.

Community

Join AI Whisperer Community!

Ready to take your AI journey to the next level? Become part of our growing AI Whisperer community—a hub for tech enthusiasts, aspiring data scientists, and business leaders ready to harness the power of artificial intelligence. Inside, you’ll find:

• Community: Like-minded individuals keen to grow together

• Curated AI Tools: Discover the latest and greatest tools to boost productivity.

• Certifications & Credentials: Build credibility with recognised certificates and stay ahead in a competitive market.

Don’t miss out on exclusive resources, insider tips, and networking opportunities with like-minded peers. Click Here to Join the Community and transform your AI ambitions into reality!

That's it for this week!
Until next time, stay curious and keep exploring the ever-evolving world of AI!
Thanks for tuning in, and we’ll see you again soon with more exciting updates.

Jul