Ad
Skip to content

Jonathan Kemper

Jonathan is a freelance tech journalist, author, and media coach. He covers open AI models, Chinese AI labs, and the technical underpinnings that make them tick – from novel training methods and scaling strategies to benchmark design. He's also an avid vibecoder.

A single prompt was enough to hijack every AI agent in an AWS account, Zenity researchers found

Zenity Labs researchers say a single publicly accessible AI agent on Amazon’s Bedrock AgentCore was enough to take over every AgentCore agent in the same AWS account and region. The attack exploited an internal AWS interface for temporary cloud credentials that agents could reach without restriction. AWS has since patched the issue and significantly tightened the agents’ default permissions.

Read full article about: South Korea bets $3.49 billion on building a homegrown frontier AI model to rival China's best

South Korea wants to develop a homegrown frontier model through government-backed equity investments of 4.7 trillion won ($3.49 billion). The funding is part of the proposed 2027 budget, though parliament still needs to approve it. A second funding track would support the adoption of Korean-built models across the country's economy.

The two finalists in the current competition between LG AI Research, SK Telecom, and Upstage won't automatically receive additional funding. Instead, the government is launching an open competition that startups can enter too. Officials have acknowledged that Korea can't compete with the largest US companies but can match leading open models from China.

When the current competition kicked off, the government pledged 530 billion won (about $390 million) to the five selected companies. The new plan roughly multiplies that figure by nine. For context, Google, Amazon, Microsoft, and Meta are planning combined investments of around $725 billion for 2026 alone, mostly for AI data centers.

South Korean industry is already pouring money into more compute capacity. In October 2025, Samsung and SK struck a deal with OpenAI to expand AI infrastructure in the country. Then in June, Samsung, SK Hynix, and the government announced a $590 billion push to scale up chip production. So far, Korean consumers have been paying mostly foreign providers for AI services. In December 2025, they spent more on ChatGPT and similar subscriptions than on Netflix.

Read full article about: CATL and Tencent back Deepseek's ballooning funding round as the AI startup eyes a 2027 IPO

Deepseek is close to raising at least $12 billion in a new funding round, Bloomberg reports. The Chinese AI startup originally aimed to raise about $7.5 billion at a valuation of roughly $75 billion. People familiar with the matter say the total could approach $15 billion. Battery maker CATL and Tencent are contributing the largest shares. The strong investor interest stems from Deepseek's V4-Flash model, which has set new benchmarks for cost and performance against Anthropic and OpenAI.

After the round closes, Deepseek plans to restructure for an IPO in early 2027. The company is building a data center with at least 160,000 AI chips from Huawei. Founder Liang Wenfeng wants to keep developing open models with freely available weights. CATL declined to comment. Tencent and Deepseek didn't respond.

Deepseek had paused the round after comments by Liang about the company's reliance on Nvidia chips went viral. In June, the company closed its first external funding round at about $7.4 billion with a valuation above $50 billion. Deepseek released V4-Pro and V4-Flash as preview versions in April and is also working on its own inference chip to reduce its dependence on Nvidia and Huawei.

Read full article about: Cohere pitches North 2 as the enterprise AI control room that works with any model

Cohere turned its enterprise platform into a control center for AI agents with North 2, handling multi-step workflows on their own and retaining context across sessions. A revamped orchestration system coordinates agents that tap into shared knowledge libraries and reusable skills while connecting to tools like Slack, SharePoint, and Jira. North 2 also generates presentations, dashboards, and simple apps straight from text prompts in the chat. Companies can run the platform on-premises, in the cloud, or fully air-gapped.

North 2 is model-agnostic. Alongside Cohere's own models like Command A+, organizations can plug in their own. Admins use "North Admin" to manage token usage, user quotas, and access rights down to individual agents. When agents take critical actions, they request human approval first. On Nvidia's Blackwell and Hopper hardware, Cohere says its models need fewer tokens. LG CNS and Bell Cyber are already running North in production.

Cohere is targeting government agencies and regulated industries that want to keep AI under their own control. That's the same audience Aleph Alpha serves. Cohere announced its acquisition of the Heidelberg-based company in April. Aleph Alpha recently resurfaced with Kolibri, a German-English language model.

 

Read full article about: ChatGPT's new ad format fills the image generation loading screen with product carousels

ChatGPT users in the US will soon see display ads while the chatbot generates an image for them. The ads are labeled as advertising, kept separate from the generated image, and designed to show products in use or the experiences tied to them. OpenAI is launching the test later this month with an initial group of advertisers.

Zwei Smartphone-Bildschirme zeigen ChatGPT während einer Bildgenerierung bei 62 Prozent. Darunter erscheint jeweils eine als "Ad" gekennzeichnete Anzeige des fiktiven Lebensmittelhändlers Heirloom Grocery mit Gemüse, Brot und einem Pastagericht, links als Bildkarussell, rechts als Reihe mehrerer Produktbilder, beide mit "Learn more"-Button.
The new display ads appear below the image ChatGPT is generating, either as a carousel or a row of product images.

OpenAI is also expanding its measurement tools. Advertisers can send conversion data to ChatGPT Ads through Hightouch, Tealium, and LiveRamp, with attribution providers like AppsFlyer, Adjust, and Branch joining the mix. Through Haus, Measured, and WorkMagic, OpenAI is testing geo-based experiments to gauge actual ad impact. On the brand safety front, the company is running pilot programs with DoubleVerify and Integral Ad Science that evaluate the ad environment without accessing real user conversations. Select advertisers can also set negative keywords to keep their ads from showing up in unsuitable conversations.

OpenAI has been running ads in ChatGPT since February, now in more than 40 countries, and has reached an annualized revenue run rate of $1 billion from ads alone. The company aims to hit $100 billion in ad revenue by 2030. According to OpenAI, ChatGPT reaches 1.2 billion people weekly, though earlier this year only about 5 percent of users were paying for a subscription.

NASA and IBM's open source lunar model turns 17 years of orbiter data into a foundation for lunar science

NASA and IBM have released the Lunar Foundation Model, one of the first open-source AI models for lunar science. Trained on nearly 2 million tile bundles, mostly from 17 years of Lunar Reconnaissance Orbiter data, it cuts the error in predicting polar ice deposits.