★ Top Story
MIT Technology ReviewAnthropic CEO Dario Amodei published a landmark essay calling for a deliberate slowdown in large language model development, citing existential risks from the technology. The move triggered a wave of responses from AI leaders and politicians — with Altman, Musk, and Hassabis loosely aligning with Amodei, while Jensen Huang pushed back forcefully. The emerging fault line between those who want to "pace the frontier" and those who see any slowdown as ceding ground to China is now the defining tension in the industry.
Read the original →
TechCrunch AIOpenAI has confirmed that it, Anthropic, and Google DeepMind have been engaged in weeks of behind-the-scenes discussions about AI safety coordination, even as the Trump administration dismisses safety concerns and prioritizes keeping pace with China. The talks add institutional weight to Amodei's public call to slow development, though critics question whether the arrangement amounts to a safety pact or a competitive cartel. The political and commercial stakes of any coordinated slowdown are enormous, with the White House signaling it will not support restraint.
Read more →
The VergeWhen the CEOs of OpenAI, Anthropic, Google DeepMind, and SpaceX loosely agreed to slow AI development, antitrust observers immediately raised red flags about whether the world's most powerful AI companies were coordinating to suppress competition under the guise of safety. The Verge examines the dual motivations at play — genuine existential concern versus the commercial benefit of freezing out smaller rivals who lack the resources to comply with self-imposed constraints. The debate cuts to the heart of how AI governance will actually function in the absence of binding regulation.
Read more →
TechCrunch AINVIDIA CEO Jensen Huang took a live call from President Trump onstage at the All-In Summit, using the moment to publicly distance himself from the Amodei-led push to slow AI development. Huang argued that AI is simply hardware and software — not an alien intelligence — and that safety is an engineering problem best left to product makers rather than regulators. His stance puts him in direct opposition to several of his biggest customers and sets up a high-profile ideological clash over the future pace of the industry.
Read more →
TechCrunch AIDoubling down on his All-In Summit remarks, Jensen Huang argued publicly that government AI regulation is unnecessary because safety can and should be engineered directly into products by their makers. He rejected the framing of AI as a novel existential threat, calling it a continuation of conventional software development. The comments put Huang at odds not only with safety-focused AI labs but also with a growing bipartisan consensus in Washington that some form of oversight is needed.
Read more →
TechCrunch AISalesforce unveiled Koa, its first CRM-native reasoning model, built on NVIDIA's open-weight Nemotron 3 Super and purpose-trained for sales, marketing, and customer support tasks. Announced at Dreamforce alongside Jensen Huang, Koa represents a new class of domain-specific models that can match or exceed general-purpose frontier models on targeted tasks — at a fraction of the cost. The launch is a direct challenge to OpenAI and Anthropic's enterprise ambitions, demonstrating that companies no longer need to rely on closed API providers to deploy powerful AI.
Read more →
Ars TechnicaA new Mozilla report, previewed exclusively by Ars Technica, finds that open-weight Chinese models have closed the capability gap with Silicon Valley's most expensive frontier models — and that paying premium prices for closed models now buys only about a four-month performance advantage at roughly five times the cost. The finding has major implications for enterprise AI procurement and for the geopolitical argument that export controls can meaningfully slow China's AI progress. It also strengthens the case for open-weight models as a viable alternative to proprietary APIs.
Read more →
The VergeA New York Times/Siena University poll of 1,503 likely voters found that 61 percent oppose the construction of new data centers to power AI infrastructure — a striking number that confirms what politicians have been responding to ahead of the midterms. The backlash is being fueled by concerns over energy consumption, environmental impact, and the siting of facilities in communities already burdened by industrial pollution. The data suggests AI's public image problem is deepening even as investment in the sector accelerates.
Read more →
TechCrunch AIA new analysis projects that the AI-driven data center boom could make the United States one of the world's largest consumers of natural gas within a decade, with domestic data centers potentially burning more gas than Germany and Japan combined by 2035. The forecast underscores the growing tension between AI's energy appetite and climate commitments, and adds urgency to debates over grid infrastructure and clean energy procurement. It also provides concrete numbers to a concern that has largely been discussed in abstract terms.
Read more →
TechCrunch AINational opposition to data center construction has reached Philadelphia, where officials are weighing a proposed facility in a neighborhood still recovering from the closure of a major oil refinery — a pattern playing out in communities across the country. Residents and advocates argue that data centers bring noise, water consumption, and emissions without meaningful local employment, repeating the extractive dynamics of the industries they replaced. The story illustrates how the infrastructure demands of the AI boom are becoming a frontline political issue in working-class communities.
Read more →
MIT Technology ReviewMIT Tech Review examines the staggering financial bets being placed on AI infrastructure, drawing on analysis from finance academics who warn that the current investment cycle carries bubble-like characteristics alongside genuine transformative potential. The piece maps out the key uncertainties — from whether AI productivity gains will materialize at scale to whether energy constraints will cap growth — that will determine whether the spending spree pays off. It's a sober counterweight to the relentless optimism of the industry's biggest boosters.
Read more →
Ars TechnicaApple has shipped iOS 27 and macOS Golden Gate 27, delivering the long-awaited overhaul of Siri that makes the assistant genuinely useful for complex, cross-app tasks for the first time. The release also refines the Liquid Glass design language introduced earlier this year and marks the final version of macOS to support Rosetta for Intel app compatibility. Early reviews suggest the new Siri meaningfully closes the gap with ChatGPT and Google Gemini for everyday AI assistant tasks.
Read more →
TechCrunch AIOpenAI has reportedly acquired Glass Imaging, a computational photography startup founded by former Apple engineers who built Portrait Mode, for approximately $300 million. The deal signals OpenAI's continued push into consumer hardware and multimodal AI, adding best-in-class camera processing capabilities to its growing device ambitions. It also raises fresh questions about OpenAI's strategic direction as it simultaneously pursues a consumer super-app, hardware partnerships, and now direct acquisitions in the device space.
Read more →
Ars TechnicaAgility Robotics has unveiled a new version of its humanoid robot with built-in collision-avoidance behaviors — including automatically stopping and lowering its center of gravity when a human enters its path — enabling it to work alongside people without physical safety cages for the first time. The capability is a significant milestone for commercial humanoid deployment, removing one of the key practical barriers to putting robots on shared factory floors. It positions Agility as a direct competitor to Boston Dynamics and Unitree in the race to make humanoids viable in real workplaces.
Read more →
The VergeMeta has launched Meta One, a tiered subscription service that bundles expanded access to its AI tools — including its new Muse assistant — with premium features across Facebook, Instagram, and WhatsApp, targeting individuals, creators, and businesses. The move represents Meta's most aggressive push yet to monetize its AI investments directly from consumers rather than relying solely on advertising revenue. Pricing and feature tiers vary, but the launch marks a structural shift in how Meta positions itself as a platform company.
Read more →
TechCrunch AIArtificial Intelligence Underwriting Company (AIUC), founded by an early Anthropic employee and the former COO of AI safety evaluator METR, has raised a $40 million Series A led by Ribbit Capital to build insurance and risk products specifically for autonomous AI agent deployments. The startup bets that as agentic AI takes on higher-stakes tasks, enterprises will need financial instruments — not just technical guardrails — to manage liability when agents go wrong. It's an early sign that the AI safety problem is beginning to spawn an entirely new financial services category.
Read more →No stories from this source today.