The pilot phase is over — that’s the verdict of 830 IT leaders in a Futurum survey that ranked agentic AI the top enterprise priority and moved the measuring stick from productivity metrics to P&L impact. The stakes for incumbent software are real enough that one widely shared analysis called it the “SaaSpocalypse” — agents replacing per-seat enterprise software. And CNBC reported Chinese AI models gaining ground inside U.S. companies for a mundane reason: comparable performance at a fraction of OpenAI and Anthropic prices.
Microsoft moved to cut its OpenAI dependency. It unveiled seven homegrown MAI models, including its first reasoning model, and tens of thousands of weekly Excel and Outlook prompts already route through them. Google’s Gemini 3.5 Pro delay kept hurting — one report tied the delay and a DeepMind talent exodus to a $225 billion drop in Alphabet’s market value (a single-source figure worth reading skeptically). OpenAI, meanwhile, shipped GPT-Realtime-2.1 for low-latency voice agents.
The constraint list grew. NTT Data reported power shortages slowing AI data-center expansion globally, and networking joined the bottleneck list as GPU clusters scale. An opinion piece counted 150,000 tech workers laid off in 2026 while companies post record profits, arguing AI is doing more justifying than automating. And Britain’s FCA-commissioned review raised the question regulators had been avoiding: whether large language models themselves need financial-services regulation.
What People Are Asking
Are US companies really using Chinese AI models?
Yes, increasingly. CNBC reports Chinese AI models are being adopted by US companies because they offer competitive performance at significantly lower cost than OpenAI and Anthropic models. CNBC →
What are Microsoft's MAI models?
Microsoft's seven homegrown AI models unveiled at Build 2026, including its first reasoning model and a coding model for GitHub Copilot — a strategic move to reduce reliance on OpenAI. Excel and Outlook already route tens of thousands of weekly prompts through them. FAQ.com.tw →
What do IT leaders say about agentic AI?
A Futurum Group survey of 830 IT leaders ranked agentic AI the #1 enterprise priority and found the pilot phase ending — success is now measured by direct P&L impact, not productivity metrics. The Daily Brief →
Why did Alphabet reportedly lose $225 billion in market value?
One report ties the drop to the Gemini 3.5 Pro delay and a talent exodus from DeepMind. The figure comes from a single source and should be read with that in mind. The Agent Report →
What is GPT-Realtime-2.1?
OpenAI's new low-latency voice agent models — GPT-Realtime-2.1 and a mini reasoning variant — offering faster, cheaper voice agents with tool use and configurable reasoning effort. MarkTechPost →
What is slowing AI data center expansion now?
Power shortages and supply chain constraints, according to NTT Data — and as GPU clusters grow, network fabric design is becoming a second bottleneck. Help Net Security →
What is the 'SaaSpocalypse'?
The argument that AI agents are starting to replace per-seat enterprise software, pressuring SaaS valuations and creating opportunities for AI engineers who build agent-based replacements. Zen van Riel →
Will the UK regulate large language models in finance?
Britain's FCA-commissioned review urges considering it, citing consumer trust issues and systemic risks from AI concentration in financial services. Insurance Journal →
Every Story From July 7
Jump to: AI labs and models (18) · AI infrastructure (12) · Policy and regulation (7) · Labor and workplace (5) · Consumer tech and culture (3) · Society, politics, and ethics (3) · Security and agent risk (2) · Markets and macro (1) · Science and research (1)
AI labs and models18 stories
Chinese AI models are increasingly adopted by US companies as they offer competitive performance at significantly lower costs compared to OpenAI and Anthropic models.
Microsoft launches a $2.5 billion unit deploying 6,000 engineers inside client companies to deploy and maintain AI systems, emphasizing on-site support and data privacy.
Google DeepMind delays Gemini 3.5 Pro to July 17 for a full architectural rebuild to better compete with OpenAI's GPT-5.6 and Anthropic's Fable 5.
OpenAI announces limited preview of GPT-5.6 models (Sol, Terra, Luna) with enhanced safety measures and government coordination before broader release.
Despite rising open source AI adoption, frontier labs like Anthropic maintain spending share as they dominate early-stage deployments, forming a two-tier model economy.
Microsoft has begun routing tens of thousands of weekly AI prompts from Excel and Outlook through its homegrown MAI models, signaling a strategic shift away from exclusive reliance on OpenAI for its 365 Copilot.
Microsoft unveiled seven MAI models at Build 2026, including its first reasoning model and a coding model for GitHub Copilot, marking a strategic move toward AI self-sufficiency.
OpenAI introduces GPT-Realtime-2.1 and its mini reasoning variant, offering faster, cost-efficient voice agents with tool use and configurable reasoning effort.
Perplexity will use Nvidia's Vera CPUs for AI agent tasks, marking Nvidia's expansion into the CPU market dominated by Intel and AMD.
Google's flagship Gemini 3.5 Pro model is delayed to July 2026 amid a talent exodus from DeepMind, wiping $225 billion off Alphabet's market cap.
Meta AI introduces Muse Image and Muse Video, featuring agentic tool use and self-refinement for advanced media generation.
This roundup analyzes June 2026 AI developments, including US regulatory struggles with Anthropic and OpenAI, model releases, and the shift from pilots to scaling.
Vivek Natarajan presents DeepMind's AI co-scientist, which uses self-play and debate to generate and refine scientific hypotheses, with promising results in biology and drug repurposing.
Lawyers can spot AI-assisted legal drafting by checking for URLs ending in 'utm_source=chatgpt.com', a simple tell that appears in court opinions and official documents.
Anthropic announces Claude updates for July 2026, expanding Cowork to web and mobile, adding Microsoft 365 write tools, and launching Claude Code and Cowork for government in beta.
Comprehensive timeline of OpenAI's GPT models and ChatGPT releases from GPT-1 to GPT-5.6, updated as of June 2026.
Explains how Perplexity AI's crawler and live search work differently from Google, and what publishers can do to earn citations.
Podcast episode discussing AI's role in accounting, including AI-powered general ledgers, data cleanup, governance, and professional standards.
AI infrastructure12 stories
A Futurum Group survey reveals agentic AI as enterprises' top priority, signaling the end of pilot phases and a shift from productivity metrics to direct P&L impact, demanding new orchestration, security, and observability infrastructure.
The author argues that AI agents are replacing per-seat enterprise software, causing a market selloff and creating opportunities for AI engineers.
NTT Data reports that power shortages and supply chain constraints are slowing AI data center expansion globally.
AI data centers face a networking bottleneck as GPU clusters grow, making fabric design and congestion control critical for AI infrastructure.
The article analyzes three major platform shifts—intelligence as a commodity, physical world automation, and geopolitical disruption—that create asymmetric opportunities for startups, with emphasis on pricing by outcomes and building defensible moats.
Microsoft's 2026 Teams update introduces a proactive AI Facilitator and UI improvements, reflecting a shift towards AI integration while raising concerns about workplace surveillance and usability.
Most U.S. factories lack automation despite high interest, with new tools and strategic shifts aiming to bridge the adoption gap.
Globant and Anthropic announce a multi-year strategic alliance to embed Claude models into AI Pods, aiming to accelerate enterprise adoption of agentic AI.
The AI in food and beverage market is projected to grow to USD 108.77 billion by 2035, driven by demand for automated quality inspection and demand forecasting.
A digest covering Anthropic's internal reasoning breakthrough, GPU commoditization, Treasury AI bubble warnings, and major infrastructure deals.
A structured guide for engineers on implementing AI-assisted coding workflows, covering tools, step-by-step processes, and productivity gains.
A vendor blog discusses how a data intelligence cloud leveraging GenAI and reinforcement learning can help financial services executives navigate regulatory complexity and modernize operations.
Policy and regulation7 stories
Britain's FCA-commissioned review urges consideration of regulating LLMs like ChatGPT due to consumer trust and systemic risks from AI concentration in financial services.
The article argues that overly broad US export controls on AI may undermine American technological leadership by pushing allies toward non-American alternatives.
Midjourney uses Hollywood's own AI usage as a legal defense in copyright lawsuits, potentially establishing industry standard for AI training on copyrighted content.
Kenya is proactively developing AI regulation through a new AI Bill, a national strategy, and cybercrime law amendments, balancing innovation with security and rights concerns.
June 2026 saw key copyright updates including USCO rulemakings, SCOTUS decision on Register's tenure, and new AI infringement lawsuits.
Attorney Ian Gottlieb apologized to the Connecticut Supreme Court for using ChatGPT to draft legal briefs that contained fake citations and quotes, in a hearing that considered sanctions for AI misuse.
The Federal Register publishes an FTC policy statement that warns against AI systems that sacrifice accuracy for ideological agendas, referencing executive orders and legal standards.
Labor and workplace5 stories
An opinion piece analyzing how profitable tech companies use AI as a justification for massive layoffs, particularly harming junior engineers, with critique of misleading AI productivity metrics.
Surveys of 50,000 students and faculty reveal that AI adoption in higher education is outpacing institutional guidance, creating a paradox where students use AI without training and faculty retreat rather than engage.
Higher education must move beyond knowledge delivery to focus on AI literacy, critical thinking, and human-centric skills to prepare students for the AI-driven job market.
At ISTELive 26, a director of learning sciences demonstrated how teachers can use ChatGPT as an 'AI brain coach' to manage burnout and improve executive function, emphasizing boundaries and guardrails.
UMaine's Summer Technology Institute offers educators training on AI and student engagement strategies for the upcoming school year.
Consumer tech and culture3 stories
Storika, an AI-native influencer marketing platform, secured seed funding from Amorepacific and others to expand in the US market.
An article providing a taxonomy of six social-first brand formats with budgeting and ROI benchmarks for creative strategists.
This article analyzes how AI features in productivity tools have evolved from content creation to context-aware assistance, aiming to reduce friction in daily work tasks.
Society, politics, and ethics3 stories
Political analyst warns that AI misuse in politics, such as spreading misinformation and fabricated content, could undermine democracy and public trust.
The article discusses how generative AI has enabled highly convincing deepfakes, creating a 'liar's dividend' and posing a serious threat to democratic processes and public trust.
The Centre for Crisis Communication warns that AI-generated disinformation and identity politics could destabilize Nigeria ahead of the 2027 polls.
Security and agent risk2 stories
A new survey reveals that most enterprises have faced AI-related security issues, with many lacking proper governance and visibility into AI systems.
Researchers warn about JadePuffer, the first fully AI-driven ransomware campaign, which autonomously conducts the entire attack chain, adapting in real time and compressing defenders' response windows.
Markets and macro1 story
North American startup funding hit a record $392 billion in H1 2026, driven by giant AI rounds like Anthropic's $65B and OpenAI's earlier round, with late-stage concentration but early-stage also rising.
Science and research1 story
This paper constructs integral tangent classes for matroids and wonderful compactifications, with the main body produced autonomously by an AI agent.
About the Rundown. The RUDI Rundown is compiled every day by RUDI from same-day reporting across the web. Every story links to its original publisher. How we build it →
Putting AI to work? RUDI helps organizations adopt AI responsibly — training, governance, and hands-on implementation. AI training · Consulting · Talk to us