AI OmniBrief – Week 34, 2025
MIT debuts AI-designed antibiotics, doctors face skill loss with AI tools, GPT-5 launches alongside Google’s Gemma, and Rolls-Royce bets on nuclear-powered data centers.
Research & Breakthroughs
MIT designs novel antibiotics with generative AI against drug-resistant bacteria
MIT researchers used generative AI to design over 36 million compounds and identified two novel antibiotics effective against drug-resistant Neisseria gonorrhoeae and MRSA in lab tests and infected mice. These compounds feature novel mechanisms disrupting bacterial membranes.
Further reading: MIT News, BBC, Cell
Study shows AI use in colonoscopies may reduce doctors' non-AI diagnostic skills
A multicentre observational study published in The Lancet Gastroenterology & Hepatology found that after routine exposure to AI-assisted colonoscopy, physicians’ adenoma detection rates during non-AI-assisted procedures decreased significantly, indicating possible skill degradation when AI was unavailable.
Further reading: The Lancet Gastroenterology, STAT News, ASCO Post
Models & Datasets
OpenAI launches GPT-5 models, restores GPT-4o picker, adds selectable speed modes
OpenAI released GPT-5 family models including GPT-5, GPT-5 mini, nano, and pro, featuring up to 256k-token context window, improved math, coding, and multimodal performance. After user backlash, OpenAI reinstated GPT-4o model picker for paid users, increased GPT-5 Thinking message limits, and introduced Auto, Fast, and Thinking modes.
Further reading: WIRED, Ars Technica, OpenAI Docs
Google releases Gemma 3 270M, an efficient open model for on-device AI tasks
Google DeepMind announced Gemma 3 270M, a 270 million-parameter decoder-only Transformer optimized for instruction following, text structuring, and fine-tuning. It runs efficiently on smartphones and edge devices, consuming 0.75% battery over 25 conversations on a Pixel 9 Pro. Available pretrained and instruction-tuned on Hugging Face, Kaggle, and Vertex AI.
Further reading: Google Developers Blog, Hugging Face, VentureBeat
Infrastructure & Hardware
Rolls-Royce plans SMRs to power AI data centers in UK and Czech Republic
Rolls-Royce aims to deploy small modular nuclear reactors (SMRs) to power AI data centers, with agreements for three SMRs in the UK and six in the Czech Republic. CEO estimates a global need for 400 SMRs by 2050, creating a trillion-dollar market. The company expects profits over £3 billion in 2025.
Further reading: BBC, Rolls-Royce SMR, MarketScreener
Tools & Platforms
Anthropic adds Learning and Explanatory Modes to Claude Code for enhanced coding guidance
Anthropic's Claude Code now includes Learning Mode, which uses a Socratic approach prompting users to contribute code, and Explanatory Mode, which describes its coding decisions step-by-step. These modes improve learning and collaboration and work with Claude Code's 1 million-token memory support.
Further reading: Engadget, Anthropic Docs, Anthropic
Parallel launches Deep Research API outperforming leading AI models including GPT-5
Parallel, founded by former Twitter CEO Parag Agrawal, released its Deep Research API to enable AI agents with advanced web research capabilities. The API supports complex multi-hop queries, delivers structured multi-page reports, and outperforms humans and AI models like GPT-5 on benchmarks.
Further reading: Parallel, Deccan Herald, YourStory
Products & Deployments
Microsoft integrates GPT-5 into Copilot with Smart Mode and expanded capabilities
Microsoft has deployed GPT-5 across Microsoft 365 Copilot, the standalone Copilot app, GitHub Copilot, and Azure AI Foundry. Features include Smart Mode which dynamically selects the best model for tasks, longer context windows, multimodal support, and free access to GPT-5 Smart Mode in the consumer Copilot app without sign-in.
Further reading: Microsoft, Microsoft Copilot Blog, Azure Blog
Apple developing AI-powered home robots, smart display, and Siri overhaul for 2026–2027
Apple is working on four AI-enabled consumer devices targeting 2026–2027 launches: a tabletop robot with a motorized arm running a new 'Charismatic' OS, a smart display with multi-user support, AI-enabled security cameras, and an upgraded Siri assistant enhancing voice-controlled app navigation. The company is also testing Anthropic's Claude as a backup AI.
Further reading: Engadget, MacRumors, eWeek
Google Gemini adds chat memory with privacy controls and Temporary Chats feature
Google Gemini AI introduces automatic memory across sessions to personalize responses by remembering key details unless users opt out. The new Temporary Chats feature allows incognito-style conversations that are deleted after 72 hours and excluded from personalization and training. These features are rolling out in select countries on the Gemini 2.5 Pro model, with plans to expand.
Further reading: Google Blog, The Verge, Google Support
ChatGPT app surpasses $2B revenue with 690M global downloads by mid-2025
OpenAI’s ChatGPT mobile app for iOS and Android has earned $2 billion in revenue to date, averaging $2.91 per install, with approximately 690 million downloads globally. Revenue surged 673% in 2025, driven by a 180% increase in downloads.
Further reading: TechCrunch, 9to5Mac, Business of Apps
Deutsche Telekom launches affordable AI phone and tablet with Perplexity assistant in Europe
Deutsche Telekom released the T Phone 3 and T Tablet 2, equipped with the integrated Perplexity AI assistant, across ten European countries. Starting at €149 and €199 respectively, devices offer built-in AI features and include an 18-month subscription to Perplexity Pro, with discounted pricing when bundled with tariffs.
Further reading: Deutsche Telekom, eWeek, GSMArena
Google launches AI-powered Flight Deals tool in Google Flights beta
Google has introduced Flight Deals, an AI-powered feature in the beta version of Google Flights. The tool uses natural language queries to help flexible travelers find cost-effective flight options from hundreds of airlines. It is available in the US, Canada, and India.
Further reading: Google Blog, Google Support, TechCrunch
LAPD tests GeoSpy AI tool for rapid photo geolocation analysis
The Los Angeles Police Department is evaluating GeoSpy, an AI-powered tool that analyzes visual clues such as soil and architecture to locate photos, sometimes pinpointing exact addresses. Licenses start around $5,000 annually, reflecting growing AI use in law enforcement.
Further reading: 404 Media, GeoSpy, Axon
Industry & Corporate
Sam Altman co-founds Merge Labs to develop brain-computer interfaces
Sam Altman co-founded Merge Labs, aiming to develop less invasive, high-bandwidth brain–computer interface technology for medical and human enhancement use. OpenAI is reportedly preparing to participate in a $250 million funding round, valuing the startup at $850 million.
Further reading: TechCrunch, The Verge, MLQ.ai
Cohere raises $500M at $6.8B valuation, names Joelle Pineau Chief AI Officer
Canadian AI startup Cohere secured $500 million in a funding round led by Radical Ventures and Inovia Capital, with participation from AMD Ventures, NVIDIA, PSP Investments, Salesforce Ventures, and others, raising its valuation to $6.8 billion. Cohere focuses on secure enterprise AI applications and appointed former Meta VP Joelle Pineau as Chief AI Officer.
Further reading: Cohere, TechCrunch, BetaKit
Microsoft offers multimillion-dollar packages to recruit AI researchers from Meta
Microsoft is actively recruiting AI talent from Meta’s Reality Labs, GenAI Infrastructure, and AI Research divisions, using competitive multimillion-dollar offers and expedited approvals, under teams led by Mustafa Suleyman and Jay Parikh, to strengthen its AI capabilities.
Further reading: Business Insider, Windows Central
xAI co-founder Igor Babuschkin leaves to form AI safety venture capital firm
Igor Babuschkin, co-founder and technical lead of Elon Musk's xAI, departed in August 2025 to establish Babuschkin Ventures, a VC firm focusing on AI safety research and startups advancing humanity. His exit follows xAI leadership changes amid controversies over its Grok chatbot.
Further reading: TechCrunch, CNBC, Yahoo Finance
Morgan Stanley forecasts $2.9 trillion data center investment by 2029, $1.5 trillion funding gap
Morgan Stanley analysts predict global AI hyperscalers will invest nearly $3 trillion in data centers by 2029, with big tech companies covering $1.4 trillion, leaving a $1.5 trillion funding gap. Major projects include Meta's Prometheus and Hyperion, xAI's Colossus, and OpenAI's Stargate, with financing increasingly reliant on private equity and debt.
Further reading: Financial Post, Sri Lanka Guardian, Duff & Phelps
Governance & Safety
Anthropic acquires Humanloop founders and team to boost AI safety and tooling
Anthropic has acqui-hired Humanloop's three co-founders and about a dozen engineers to enhance enterprise AI tooling focused on safety, iteration speed, and developer experience. Anthropic offers its Claude AI to all U.S. federal agencies for $1 per agency annually, facilitating government AI adoption.
Further reading: TechCrunch, Anthropic, GSA.gov
US covertly tracks AI chip shipments to prevent illegal diversions to China
US authorities have secretly placed location trackers in select shipments of Nvidia and AMD AI chips and servers to monitor unauthorized rerouting to China under export controls. Former President Trump also authorized a deal requiring Nvidia and AMD to pay 15% of China sales revenue for limited chip exports.
Further reading: Reason, BBC, Fortune
Meta’s internal AI policies allowed inappropriate chatbot interactions with minors, prompting US Senate probe
Leaked internal Meta documents revealed AI chatbots were permitted to engage in romantic or sensual conversations with users identifying as minors. The revelations led over 100 US senators, including Senator Josh Hawley, to call for a congressional investigation. Meta has removed the contentious policy portions and states they prohibit sexualized interactions with minors.
Further reading: BBC, The Guardian, CNBC
US senators launch congressional probe into Meta’s AI chatbot policies over child safety
Over 100 US senators, led by Senator Josh Hawley, have called for a congressional investigation into Meta's AI chatbot policies after reports revealed internal guidelines allowed chatbots to engage in romantic and sensual conversations with minors.
Further reading: Hawley Senate, The Guardian, U.S. News
Geoffrey Hinton warns companies prioritize profit over AI safety amid superintelligence risks
AI pioneer Geoffrey Hinton warns that tech firms prioritize short-term profits over long-term AI safety. He emphasizes risks from potential superintelligent AI systems gaining control and calls for improved alignment mechanisms and stronger governance.
Further reading: CNN, CBS News, MIT Sloan
You’re reading AI OmniBrief - the weekly AI newsletter for executives, engineers, researchers, and just anyone who prefers concise briefs over scattered updates.
Presented to you by Matthias Isler, Fractional CTO and AI advisor. I help technology ventures make AI real: from strategy and integration to building teams and products.
AI OmniBrief is free today. If you found this valuable, you can support it by pledging a future subscription. You’ll only be charged once payments are enabled.