AI Flash · live feed
A timeline of everything that has passed through the radar — newest first, with an evidence status for each item.
Continuously updated
1096 items in the selected feed
How to read evidence statuses ↓
Virginia Governor Abigail Spanberger issued an executive order that prohibits NDAs for data center projects, introduces a review of their approval process, and establishes an AI task force to address risks such as job losses and data protection.
Disney has created the position of CTO for the first time in its history and appointed Karandeep Anand, the former CEO of startup Character.AI – a company that Disney accused in September 2025 of infringing copyright on its characters.
Google updated the experimental agent CC (formerly Daily Brief) — it now handles specific household tasks such as signing permission slips, shopping lists and meal plans, runs on the Antigravity framework with Gemini models and supports up to 6 family members in the USA.
Kimi K3 (2.8 trillion parameters) from Moonshot AI has also been available on the Amazon Bedrock platform since 18 September 2026, with support for prompt caching and AWS data protection. The model was released in July 2026 as the first open model in the 3 trillion parameter class.
The Verge contacted Hollywood studios (Disney, Netflix, Amazon, Lionsgate) for comment on warnings about the existential risks of AI. The studios did not respond, but the unions SAG-AFTRA and WGAE responded with a call to focus on problems already occurring in the industry.
AWS described the migration of a healthcare AI agent orchestrating three models (BioM-ELECTRA, Llama 3.1 70B) from self-managed ECS/Fargate to the managed runtime Amazon Bedrock AgentCore without requiring the agent code to be rewritten.
Editor Emanuel from 404 Media describes months of research into the misuse of AI-generated music on Spotify that appears under the names of real artists. According to him, the problem continues unabated, and hundreds of examples are easy to find.
TechCrunch Disrupt 2026 (13–15 October, San Francisco) hosts a panel by Nvidia on choosing between open and proprietary AI models for startups. According to Nvidia, its open Nemotron models were cited by 145 papers at ICML 2026; the open model Nemotron 3 Super has 120 billion parameters.
Super PACs linked to Anthropic, OpenAI and the investment firm Andreessen Horowitz have put approximately 1 million dollars into the Senate race of Republican Senator Mike Rounds in South Dakota ahead of the November elections, according to Wired.
Anthropic says that Claude leads 26 % of the work on developing its own models (level AL4), compared with less than 1 % in February 2026. However, agreement of only 33% between human evaluators and 59% between the model and humans calls the methodology into question.
Google is expanding AI & Economy Research Program with new senior researchers – economist Philippe Aghion (Nobel Prize in economics 2025) as an advisor, Ajay Agrawal as a Visiting Fellow, and Anu Madgavkar together with Daniel Rock as directors of research into the impacts of AI on work and productivity.
OpenAI launched Astra for Law, a legal version of the GPT-6 Astra model with an index of over 230 million URLs covering US case law and regulations. In the Legal Research Bench test, it achieved a 54 % success rate compared with 38.7 % for the standard GPT-6 Astra with web search.
Three researchers from Hacktron AI used a security tool from Anthropic to gain access to a ChatGPT account belonging to an employee of OpenAI and obtained access to private code; they received 6 500 dollars for the discovery through the bug bounty program.
AWS introduced Amazon SageMaker HyperPod Inference Gateway, a Kubernetes addon for GPU-aware routing of LLM requests. According to the company, it reduces time to first token by up to 82 %, in one example from 4.4 s to 800 ms, without requiring changes to applications.
Google, together with designers Jane Wade and Sergio Hudson, created Styling Suite for virtual styling and Runway Visualization for simulating a fashion show in Google Flow, with the tools deployed at New York Fashion Week.
A librarian at University of Virginia proposes dividing answers from AI systems into four types – factual, interpretive, constructive and strategic – with a different verification process for each.
OpenAI announced Australian Youth Safety Blueprint – a six-stage plan for safer and more empowering AI experiences for young people. The source does not specify the content of the pillars.
Scientists from UCSD, NASA and the US National Weather Service are developing TACLS software, which uses satellite data and machine learning to target earlier detection of flash flood risk. It is being piloted in California, with expansion to all NWS offices planned.
Experts from the US (Melanie Sisson, Brookings) and China (Tianjiao Jiang, Fudan University) proposed rules ahead of the Trump-Xi meeting on 24 September prohibiting AI from autonomously launching nuclear weapons or attacking nuclear command systems. Binding mechanisms are still lacking.
Napster, after pivoting from a music streaming service to an AI platform, is launching digital AI personas of teachers with Gems Education in Dubai, trained on their materials and intended to provide students with round-the-clock support.
Mustafa Suleyman, head of the artificial intelligence division at Microsoft, warned that without sufficient safety measures, AI could lead to the emergence of a “silicon species” competing with humans. BBC reported the news.
The European Commission has announced Data Union Conference on the future of EU data policy, the availability of high-quality data for AI and the simplification of digital rules. The event will be streamed live, and attendance is subject to approval.
Writer Leoš Kyša predicts that within a few years or months, AI will write books better than 80 % of human output. According to him, fans are already using AI to generate unofficial sequels to his book series.
A research team released a description of the synthetic STEM dataset QVAC Genesis III (191.43 billion tokens, 19 domains). According to the authors, on models with 1.7 billion parameters, it outperformed Cosmopedia-v2 and Cosmo-1B on the ARC, GPQA Diamond and MMLU STEM benchmarks.
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
