Archive · month
Nvidia unveiled the RTX Spark chip (an Arm-based SoC with Grace CPU and Blackwell GPU) for local agentic AI at IFA 2026 in Berlin. Lenovo, Dell, Asus and Microsoft are preparing the first laptops; pricing is not yet known.
Sam Altman, CEO of OpenAI, described the boom in building AI compute infrastructure as “unsustainable silliness” and criticized neocloud providers building capacity without corresponding demand. He described the expansion of OpenAI itself as profitable.
In June 2026, the German government decided to establish the federal institute DE-AISI to analyse the safety of advanced AI models. Four cities are competing to host it, while Bitkom and DFKI are disputing the scope of its mandate; UK, USA and nine other countries or the EU already have similar institutes.
Meta has ended employee evaluations based on AI tool adoption (dashboards, token counts) and is returning to evaluations based on work results. At the same time, it is testing the agentic tool Hatch, with employees raising privacy concerns about connecting it to personal accounts.
Palo Alto Networks acquired the startup Console, which uses AI agents to automate IT helpdesk tasks, for 500 million dollars. This is the seventh acquisition by the company in 2026, and Console will be integrated into the Cortex platform.
Startup Mostik has developed a method through which AI models communicate directly using mathematical values in their weights, without text output. Combining GLM-5.2 (753 billion parameters) with Qwen-3.5 (4 billion) produced a hybrid system at one fifth... one twentieth of the price of the full GLM, with performance exactly between…
AWS described a solution using the Amazon Bedrock AgentCore service that uses autonomous agents to analyze .NET code and automatically generate architectural diagrams and documentation integrated into CI/CD. It has been running in production since the 1st quarter of 2026 at a global financial firm focused on interdealer broking.
GitHub has made content exclusions available (GA) in the GitHub Copilot app and Copilot CLI. Files excluded by administrators at the enterprise, organization, or repository level will not be used as AI context. Available for Copilot Business and Enterprise.
The Johnson County Sheriff's Office in Texas had the AI tool Axon Draft One write part of a police report about a case in which it searched for a woman after a self-managed abortion using a network of over 80 000 Flock cameras. According to the DA, no crime had been committed; the investigation was prompted by her abusive partner, not her family.
Google has expanded its multiyear agreement with MrBeast: in addition to videos about surviving in extreme conditions with Gemini (premiere on 5 September), the creator will also appear in a separate advertising campaign for Gemini. Context: according to Bloomberg, YouTube is attracting creators with exclusivity because of competition from Netflix.
New: MrBeast will appear in the own Gemini campaign run by Google (expanding the scope of the partnership); The partnership is part of efforts by Google to compete with Netflix in content; YouTube offers creators exclusive agreements to keep content on its platform; A more detailed description of the content mechanics: teams race to navigate three climate zones using Gemini to identify hazards and survive
A research team introduced the ARISE-RL framework for training agents using RL with rubric-mediated co-evolution of Generator and Solver, complemented by the RG-SED method and ECR-Bench benchmark; according to the authors, it achieves state-of-the-art results.
A research team has released the RestoreBench benchmark for testing LLM agents (chatbot, single agent, multi-agent) in diagnosing and correcting non-convergent power flow cases in two electrical grids with 46 cases per grid. The code is publicly available.
According to the LA Times, Google is in talks with Disney, Warner Bros. Discovery and Universal about paid licenses to train Gemini AI models on their content—up to 40 million dollars for one character. The agreements have not yet been finalized.
AI training data startup AfterQuery has reached a valuation of 3.2 billion dollars, according to Forbes, ten times its valuation in the April Series A round. Y Combinator calls it the fastest path to unicorn status in its history.
NVIDIA and CrowdStrike announced CrowdStrike SafeMind, an agentic cybersecurity system focused on automated defense against automated attacks, at the Fal.Con 2026 conference in Las Vegas.
GitHub has launched a feature in public preview that allows Copilot Code Review to approve pull requests automatically. It is disabled by default and can be enabled at the enterprise, organization, or repository level.
Google DeepMind head Koray Kavukcuoglu said that the Gemini 4 model is in the fine-tuning phase and the company plans to release it much earlier than by the end of the year. He previously said that leadership in frontier AI is the company's sole priority; the Gemini 3.5 Pro model, despite being announced, was never released.
New: Gemini 3.5 Pro was never released, even though it was announced in May 2026; the project missed 3 deadlines; Demis Hassabis stepped down from leading Google DeepMind in August 2026; Kavukcuoglu replaced him; Kavukcuoglu explicitly rejects AGI as a priority, calling it 'not the right conversation'; Google DeepMind operates under competitive pressure as a 'Gemini factory' (a model-production facility); Gemini 4 is in the 'early phase of post-training', with release planned 'much earlier' than by the end of the year
Startup Empirik has become independent after three years of incubation at Sequoia Capital and raised 21 million dollars in seed funding from Sequoia, Canapi and Alumni Ventures for a tool intended to autonomously predict infrastructure outages and review risky system changes.
Jamf published a production serverless system that tracks individual engineers' daily spending on Amazon Bedrock and gradually blocks more expensive models (Claude Opus at 80 %, Claude Sonnet at 100 % of the budget) without interrupting ongoing work.
ZS Associates built a secure multi-tenant platform for ad-hoc analytics on Amazon SageMaker Studio in a regulated healthcare environment. It operates 200+ SageMaker domains; the source gives conflicting user counts (both 1 000+ and 500+ per day).
AI Radar monitors Czech and international sources every day, looking for changes that truly deserve attention.
The result is a quick overview of what has actually changed in the AI world, rather than another stream of articles.
Use the CS/EN switch to read the same Radar in Czech or English. English content is published after its translation has been checked, so new and older items may appear later.
Each row represents one event — not one article. At a glance, you can see its significance, credibility and main point.
Illustrative example, not a current news item.
Agent mode is available to all paying users
Until now, the mode was available only on the highest plan; it is now available on all paid tiers without a waitlist.
The detail page contains a fuller summary, its significance and original sources. Practical impact appears in the detail and the For individuals and For businesses views. An AI Flash item reaches the main selection only after it has been expanded and meets the publication rules.
We first summarize each event in the same way for everyone. Based on those same facts, we then explain what the change means for your own use and what it could mean for how a company operates.
Today’s briefing is the same for everyone. Pages For individuals and For businesses can be found in the main navigation — they select only events relevant to the given use case.
Reset history · Outage history · Subscribe in a feed reader (RSS) · Methodology and label explanations
