Archive · month
FCC banned imports of new Chinese humanoid robots and voltage converters in late July 2026 on national security grounds. China (Unitree, Agibot) dominates the market; analysts expect a strategic rather than immediate impact and pressure to diversify suppliers.
Anthropic and AWS have published a reference architecture for deploying Claude apps gateway – a management layer for Claude Code and Claude Desktop – on AWS Fargate with centralized OIDC authentication, group-based model access control and cost tracking via OTLP.
OpenAI announced that it had resolved an incident involving increased error rates in ChatGPT Work; the sources differ on whether Workspace Agents (Agent) or Conversations were also affected. All affected services have now been fully restored.
New: The incident had the status 'Investigating' during the investigation; During the investigation, the ChatGPT Work component had the status 'Degraded performance'; OpenAI was actively investigating the issue involving increased errors; The new article captures the phase of the incident BEFORE its resolution
NVIDIA claims that scaling AI computing is constrained not just by overall performance, but by the way electricity gets from the grid to the GPUs — according to the company, traditional alternating current (AC) distribution is becoming a bottleneck.
NVIDIA has published a technical description and benchmark results for the Nemotron 3.5 Super (120B/12B active) and Lightning (30B/3B active) models. The Lightning model itself was released by the company as open source under the OpenMDW 1.1 license back on 11 August, together with the NeMo Switchyard routing tool.
New: Lightning is a MoE hybrid Mamba-Transformer model with 3B active parameters out of 30B total; Super is a 120B model with 12B active parameters, multimodal, intended for code engineering; Both models include long-context training and native speculative decoding; The post-training pipeline uses RLVR, RLHF, and for Super, multi-teacher on-policy distillation (MOPD); Super achieves 51.78% on SWE-bench Pro and has 4.6× fewer active parameters than Nemotron 3 Ultra
According to Spotify, more than 50 million tracks identified as low-quality AI-generated content or spam were removed from recommendations over the past year. The company also reports a significant decline in listener complaints about this content in recommendations.
New: Spotify removed more than 50 million tracks, including AI slop and spam, from recommendations over the past year.; Spotify says listener complaints about this content fell significantly following changes to recommendations.
From 26 August 2026, OpenAI is expanding ad testing in ChatGPT to 31 European countries (Germany, France, Spain, Italy and others). Only users of the Free version and the Go plan at 200 Kč/month will see ads.
New: Expansion of testing to 31 European countries planned for 2026-08-26; Ads are shown only to users of the Free version and the Go plan (200 Kč/month); Plus, Pro and Enterprise users are protected from ads; 9 named countries included: Germany, France, Spain, Italy, Sweden, Norway, Denmark, the Netherlands, Austria; Ad Manager (Ads Manager) for Europe will be available later in the summer
Anthropic launched an API that allows regulators, media outlets, fact-checkers and companies to verify an invisible watermark in text from Claude — a response to EU AI Act obligations in effect since August 2026. Critics point out that rewriting removes the watermark from the text and that the method may affect quality.
New: Anthropic launched a watermark verification API for verifying watermarks; The API is available to regulators, media outlets, fact-checkers, nonprofit organizations and businesses; The system is based on the SynthID method from Google; The API allows eligible entities to verify the presence of watermarks in text from Claude
A research team released the SkillReason framework and SkillReason-Bench benchmark (3729 queries, 61228 skills, 9 domains) on arXiv for more accurate retrieval of AI agent skills from vaguely worded requests.
Journalist Casey Newton (Platformer) criticizes Mark Zuckerberg’s essay on superintelligence for overlooking risks, including job automation. He compares superintelligence to a dragon – a visible but hard-to-imagine power.
NVIDIA has made the open-source model Nemotron 3.5 Lightning (30B parameters, 3B active per token, MoE) available in Ollama for local AI agents. It supports a context of up to 1M tokens, runs on NVIDIA GPUs as well as Apple silicon and can be further trained for custom tasks.
According to procurement documents, ICE will pay LexisNexis 6.7 million dollars for data (LexID, Accurint) linked to systems from Palantir, including AI identification of people and facial recognition for deportation operations.
According to The New York Times, Amazon is financing a gas-fired power plant in Texas for an AI data center that could become the largest source of climate pollution in the USA. This is raising concerns about community health and drawing criticism over the bypassing of environmental assessments.
GitHub released Copilot SDK for Java in version 1.0.7-preview.1 as a Maven dependency. The SDK is framework-agnostic (with no dependence on Langchain4j or Spring AI) and, thanks to BYOK, supports any models, including OpenAI, Anthropic or Azure, without requiring a Copilot subscription.
Sarah Friar, CFO at OpenAI, describes five lessons from building an AI-native finance department – automated forecasting, stronger controls and measurement of ROI in AI.
A growing number of AI apps and websites scan faces and assign them numerical attractiveness scores, or compare users in real time. According to research, they are linked to looksmaxxing culture and increase the risk of body image disorders, muscle dysmorphia and eating disorders among young people.
OpenAI paused part of the development of the forthcoming model Astra after an internal review under Preparedness Framework showed that the model had reached the critical cybersecurity threshold and can independently develop zero-day exploits. The model is unrelated to the attack on Hugging Face.
OpenAI announced that the incident involving an increased error rate in the Conversations service has been resolved. According to the company, the affected services have fully recovered and the component is now in Operational status.
An analysis on Tech Policy Press (authors Neumann, Sargeant, Singh) argues that AI safety regulation based on the text of system prompts is insufficient – prompts provide only weak assurances about behavior and can conceal the gap between intent and actual model output. Evidence: incidents involving Grok and GPT-4o.
Meta Platforms launched the open-source model Muse Glimmer, a smaller alternative to models from competitors, designed to run directly on users’ devices with consumer graphics cards, without needing a cloud connection.
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
