AI-generated
Tuesday, 22 September 2026 | 53 articles
Alibaba unveiled plans for a model with up to 10 trillion parameters alongside a new chip meant to power 20GW of data centers by 2032, while Anthropic, OpenAI, and Google all reported security breaches affecting their AI models. Meanwhile, Amazon blocked Meta's Muse AI shopping agent even as Shopify partnered with it for automated checkout, highlighting growing friction over agentic commerce.
TypeSafe AI has introduced Jev, an AI judgment model it says is 75 times faster and 170 times cheaper than OpenAI's model for classification tasks. The company positions Jev as a specialized alternative aimed at cutting operational costs for AI-driven decision-making. Source: chosun.com
Jev, a new AI model from TypeSafe AI, is drawing attention among developers for its distinct approach to handling decision and judgment tasks. It's being framed as a 'generalist' model that could rival established systems in speed and efficiency. Source: indianexpress.com
Major AI companies including Anthropic, OpenAI, and Google's Gemini have each faced recent security incidents involving their large language models. The repeated breaches raise questions about why current safety defenses for LLMs keep getting bypassed. Source: eu.36kr.com
Researchers from Georgia Tech, Nvidia Research, and Stanford built BOOST, a runtime system that taps host memory and HBM at the same time to accelerate LLM inference. The approach aims to reduce bottlenecks that typically slow down large-scale model serving. Source: indexbox.io
Businesses are shifting quickly from testing generative AI to finding concrete use cases that generate measurable enterprise value. Consultancies now focus on helping organizations turn large language model capabilities into practical, revenue-driving applications. Source: tynmagazine.com
As AI adoption grows, more users are turning to offline, on-device models that keep prompts and data away from third-party servers. After testing a local LLM on a smartphone for five practical tasks, the experience proved it can be a genuinely useful privacy-focused alternative. Source: makeuseof.com
Anthropic has teamed up with more than a dozen industry partners to bring its Claude AI models into the workflows of registered investment advisors. The move could test how eagerly tens of thousands of financial professionals adopt AI-assisted tools in their daily practice. Source: financial-planning.com
Amazon Web Services introduced Strands Harness, an open-source tool designed to help developers scale AI agents across different cloud environments. The launch addresses a common pain point in deploying agents beyond a single infrastructure. Source: siliconangle.com
Amazon has cut off access for Meta's new Muse AI agent on Amazon.com, citing data security concerns. The move deepens tensions between the two tech giants over autonomous shopping agents. Source: geekwire.com
Shopify will let Meta's Muse AI agent complete purchases on behalf of shoppers across its network of online stores. The deal marks a contrast to Amazon's more cautious stance on agentic commerce. Source: wsj.com
NVIDIA's technical blog discusses how to test whether an AI agent can reliably complete complex, multi-step tasks in real environments. The focus is on measuring chains of tool calls rather than isolated responses. Source: developer.nvidia.com
A report on Jev AI points out that many enterprise agent pipelines waste full LLM calls just to route simple decisions, like choosing which tool to use. This inefficiency also opens the door to prompt injection vulnerabilities. Source: venturebeat.com
Muse is designed to autonomously handle tasks using a user's email, payment details, and online accounts. Meta's biggest challenge now is convincing people to trust it with that level of access. Source: fastcompanyme.com
Amazon cut off access to Meta's Muse AI agent on its shopping site after Meta declined a request to withdraw it. Multiple media sources across the political spectrum have reported on the escalating standoff. Source: newscord.org
Amazon has blocked its shoppers from using Meta's newest AI agent, Muse, on its e-commerce platform. The restriction reflects ongoing friction over third-party AI agents operating within Amazon's ecosystem. Source: mediapost.com
Meta's Muse AI agent racked up 730,000 downloads in its first five days and 2.5 million within thirteen days of launch. The rapid growth puts it in direct competition with established AI assistants like ChatGPT, Grok and Claude. Source: cnbc.com
Meta shares jumped 11% on Monday as excitement grew around its new Muse AI agent app. Still, analysts caution that early buzz doesn't ensure the product's long-term success. Source: taekim.substack.com
The public resignation of Anthropic's Jacob Coxon has intensified debate over AI safety practices. The episode highlights key considerations CISOs should weigh when deploying AI within their organizations. Source: techtarget.com
Ahead of a planned meeting between President Trump and Chinese President Xi Jinping, the two countries reportedly agreed on a mechanism to notify each other about AI safety concerns. The move signals growing bilateral attention to managing AI risks. Source: aimagazine.com
US Treasury Secretary Scott Bessent said talks with China on trade and AI safety have been productive. The negotiations come as concerns mount over the potential dangers posed by advanced AI systems. Source: dw.com
Concerns over political and existential risks tied to AI have pushed at least three Republican senators to start taking legislative action. The shift reflects growing public attention to AI safety issues ahead of upcoming elections. Source: politico.com
In an interview, AI risk researcher and author of "Superintelligence" Nick Bostrom describes himself as a "fretful optimist" about the technology's future. He argues superintelligence represents a critical threshold humanity will eventually have to cross. Source: english.hani.co.kr
An opinion piece suggests that existential AI risks and disputes over data center locations and regulation could be addressed through a structured mediation process. The author argues these two issues share common negotiation dynamics. Source: mediate.com
A commentary argues that the race for AI growth reflects a broader obsession with expansion at any cost. It claims a small group of tech leaders is effectively gambling with humanity's future in pursuit of progress. Source: resilience.org
Warnings about uncontrolled AI development are adding pressure to the upcoming talks between President Trump and Chinese President Xi Jinping. Despite the urgency, significant obstacles remain to reaching concrete agreements. Source: abc27.com
NVIDIA has pushed back against claims that AI poses existential threats to humanity, amid an ongoing industry debate on safety. While some experts continue to warn of serious dangers, the company downplays such concerns. Source: hi-tech.ua
A newly disclosed flaw affecting Claude Code, Codex, GitHub Copilot, and Gemini CLI allows a malicious repository owner to silently swap out a pinned plugin for harmful code. The attack requires no user interaction, making it especially dangerous. Source: thenextweb.com
At pladis, staff previously spent hours every week on routine work like drafting content, summarizing documents, and compiling reports. The company reports that active Copilot users now reclaim more than 12 hours monthly thanks to AI-assisted automation of these tasks. Source: microsoft.com
The Plugin4Shell vulnerability bypasses SHA-based plugin pinning by creating a branch named after the hash itself, affecting Claude Code, Codex, GitHub Copilot, and Gemini CLI. So far, only two of the four affected tools have received a fix. Source: pasqualepillitteri.it
Microsoft completed a large-scale engineering project driven by AI agents, migrating GitHub Copilot's public runtime from TypeScript to Rust. The effort cost around $120,000, produced roughly 800,000 lines of Rust code, and delivered a performance improvement of nearly 16 times. Source: finance.biggo.com
Taptwice Media, an AEO agency based in India, published an initial visibility study showing significant inconsistencies in how leading AI engines present and recommend industrial quoting and CPQ software vendors. The findings highlight uneven brand representation across different AI platforms. Source: m.thewire.in
Toronto-Dominion Bank (NYSE:TD) continues its US growth push, targeting 100 branches in Florida and the Carolinas by 2028 while navigating regulatory hurdles. The bank is also expanding use of AI for anti-money-laundering monitoring and customer call handling. Source: tradingview.com
OpenAI plans to retire Sora 2 on September 24, 2026, just two days after a fresh comparison against Veo 3.1 and Kling 3.0 was published. The looming shutdown adds urgency to an otherwise routine AI video generator showdown. Source: tech-insider.org
Alibaba's Qwen team has open-sourced Qwen-Image-2.1, a model that unifies text-to-image generation and image editing into one workflow. It aims to let users both create and refine images without switching tools. Source: technode.com
Alibaba's Qwen team released Qwen Image 2.1 with fully open weights, prompting immediate hands-on testing by creators. The video walks through what the model can generate and how it stacks up against closed alternatives. Source: youtu.be
Google is embedding its latest video generation model directly into Google Ads, letting advertisers produce campaign-ready video clips without external tools. The move aims to streamline video asset creation for ad campaigns. Source: almcorp.com
Building a marketing campaign often requires multiple connected assets — website imagery, social graphics, product shots, and a short video. A workflow pairing Kling 4.0 with GPT Image 2.5 shows how AI tools can generate all these pieces consistently. Source: appleworld.today
From local newsletters to climate campaigns, audiences increasingly expect video content rather than plain text. AI-powered video tools are now making it easier for independent creators and small outlets to meet that demand without big production budgets. Source: ipsnews.net
Gener8ion's music video "Storm" won early praise from fans including Charli XCX and Josh Kushner. Soon after, its distinctive look was widely imitated across AI-generated content. Source: insidehook.com
Figure has committed billions of dollars toward gathering more training data for its humanoid robots. Just four weeks after the announcement, the company reports that the expanded dataset already boosted its robots' success rate at household chores by a factor of six. Source: forbes.com
Machine learning models trained in one environment often struggle when deployed in a different, real-world setting. Researchers are now using CLIP to help classifiers adapt to new conditions without needing access to the original training data. Source: bioengineer.org
As data privacy regulations tighten worldwide, AI systems are increasingly required to erase specific information they've learned. A new hypernetwork-based framework allows models to forget targeted data without needing access to the original training set. Source: bioengineer.org
According to Oura's IPO filing, the wearable ring maker's public offering is valued at $2.2 billion, but much of the proceeds will go to existing shareholders rather than the company itself. Forerunner Ventures alone plans to sell its full stake for up to $1.26 billion. Source: techcrunch.com
Pave Finance raised an oversubscribed $15 million Series A round at a $100 million pre-money valuation. The company said assets under management by advisors using its AI-driven portfolio tools have now exceeded $130 billion. Source: pulse2.com
An overview of the largest venture capital deals in the US during August 2026, based on CrunchBase data. The list highlights which startups attracted the most investor attention that month. Source: alleywatch.com
Defense-tech startup Mach Industries closed a $600 million extension of its Series C round, pushing its valuation to $3.7 billion. The funding will help scale up its military manufacturing capabilities. Source: latimes.com
Mind, a cybersecurity startup based in Seattle, plans to at least double its 70-person team after closing a $72 million Series B round led by Crosspoint. The fresh capital will fuel hiring and product expansion. Source: bizjournals.com
Rainmaker secured a $100 million Series B round to expand its research into low-cost water production and weather modification. The company says it's behind the first scientifically validated case of manmade precipitation. Source: prnewswire.com
Alibaba Group is launching what it describes as China's most powerful AI chip, an accelerator aimed at rivaling Nvidia. The chip is meant to support the company's plans for massive data-center capacity growth by 2032. Source: bloomberg.com
Alibaba CEO Eddie Wu said the company plans to train a new AI model with between 5 and 10 trillion parameters. The announcement came alongside the unveiling of a new AI accelerator chip. Source: reuters.com
Safety for physical AI systems covers hardware, software and rigorous validation. NVIDIA's Halos platform aims to deliver full-stack safety for autonomous vehicles and robots. Source: blogs.nvidia.com
Alibaba presented a new AI chip on Tuesday, describing it as the most powerful ever developed in China. The internet giant is positioning the chip as part of its broader push into AI infrastructure. Source: investing.com
NVIDIA's new qualification program helps builders of AI data centers identify power and cooling products that meet its DSX specifications. The initiative starts with battery energy storage systems and is expected to expand to other infrastructure components. Source: blogs.nvidia.com
AIskimIQ is an independent project. If you find it useful, you can support its development with a coffee.
Published by Martin Ševčík
22 September 2026 at 05:09
The most telling story today isn't about a model at all—it's about who gets to talk to whom. Amazon blocking Meta's Muse shopping agent while Shopify opens its doors wide to the same agent tells you everything about where the real battle for AI's next decade is happening: not in benchmarks, but in access control.
Think about what's actually going on here. Muse is an AI agent designed to shop autonomously on your behalf—browse, compare, click "buy." Amazon says it's blocking Muse over data security concerns, which may well be genuine, but it's hard not to notice that Amazon has its own agentic shopping ambitions and zero interest in letting a Meta-built intermediary sit between it and the customer relationship it's spent decades building. Shopify, by contrast, has nothing to lose and plenty to gain by welcoming any agent that drives checkout volume through its merchants. This is the same dynamic we saw with browsers and app stores twenty years ago, just replayed with AI agents as the new interface layer. Whoever controls the last mile of a transaction controls the leverage, and platforms are already drawing battle lines before agentic commerce has even matured. Expect more of this—not fewer walls, but more selective ones, as every major platform decides which agents it trusts enough to let transact on its turf.
Meanwhile Alibaba is making a very different kind of bet, and a much bigger one. Eddie Wu announced plans to train a model with 5 to 10 trillion parameters, alongside a new in-house AI chip meant to underpin 20GW of data-center capacity by 2032. That's not an incremental scaling exercise—that's Alibaba trying to build the entire stack, from silicon to frontier model, largely independent of Nvidia and, implicitly, independent of US export controls. Whether a model at that scale delivers meaningfully better capability than something a tenth its size is genuinely unclear; parameter count alone stopped being a reliable proxy for quality years ago. But the chip is arguably the more interesting signal. China's AI companies aren't just training around GPU shortages anymore—they're building for a future where they assume the shortages get worse, not better. That's a geopolitical hedge as much as a technical one.
And then there's the security story that deserves more attention than it's getting: Anthropic, OpenAI, and Google have all disclosed breaches or bypasses in their models recently. When it happens to one lab, it's a vendor problem. When it happens to three of the most well-resourced labs in the world within the same stretch, it starts looking like a structural one. Guardrails built through fine-tuning and RLHF keep getting treated as permanent fixes when they're really just speed bumps against a determined adversary. By the way, this is exactly the kind of thing that gets glossed over in the rush to ship agentic features—if your model can be jailbroken, what happens when that model also has a credit card and checkout access?
If agents are going to shop, negotiate, and transact on our behalf, the security question stops being academic. I'd rather see labs slow down and get this right than watch it become next year's headline for the wrong reasons.