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 ↓
The Hard Fork podcast (New York Times), which focused on AI safety and LLMs after the launch of ChatGPT, is ending. Creators Casey Newton and Kevin Roose are launching a new podcast, Machine Gods, in partnership with NPR.
Swedish startup Scaleout Systems is adapting smaller computer vision ML models to run on edge hardware in drones and at field stations, and was selected for the NATO DIANA 2025 accelerator.
GitHub has expanded the usage metrics API for Copilot CLI with new fields tracking the use of skills, custom agents, MCP servers, slash commands and plugins – administrators can now see which agentic tools developers actually use.
Google launched the Learning Interactives pilot program, which uses generative UI to enable teachers to create AI-generated interactive simulations for STEM subjects. The library contains over 30 materials, and schools using Google Workspace for Education have access.
The Australian Fair Work Commission requires mandatory disclosure of generative AI use in submissions from 20 October 2026. A breach may result in a document being given less weight, payment of the opposing party’s costs or dismissal of the claim.
According to AI Now Institute, an emphasis on existential AI risks in public debate may crowd out specific regulatory proposals. The institute calls for an analysis of specific risk mechanisms before addressing scenarios of human extinction.
Legal experts warn that the words “slowdown” and “pause”, which AI companies use in connection with safety initiatives, may be interpreted under the Sherman Act as an anticompetitive agreement to restrict output, rather than as a legitimate safety measure.
Since September 2025, Alight Solutions has been beta testing a tool from Phenom to detect fraudulent applicants; the agent detected a candidate applying twice under a different identity. The company sees the benefit primarily as time savings, rather than direct financial savings.
Part of an overview of multiple AI topics; only this event has been covered.
According to audit records, Flock Safety operated a fake account called “Flock City PD” that ran sensitive searches such as “Star of David” or stickers featuring Trump on live license plate reader cameras during sales demonstrations.
Amazon launched Amazon Connect Talent, a tool that has AI agents conduct candidate interviews and assessments for hiring at scale (retail, logistics, hospitality); recruiters set the criteria and make final decisions based on a dashboard with scores and transcripts.
Baseten, through its research division Base Labs, has launched a partnership with Hugging Face and Goodfire AI to build safety infrastructure for open models – a response to the growing number of models stripped of safeguards using abliteration.
Pinterest is testing the Restyle feature in the USA and Canada, using AI to let users visualize interior changes — furniture, colors, styles — directly on a photo of their own room.
A commentary on Tech Policy Press criticizes how belief in the imminent emergence of AGI/ASI drives investment narratives at tech companies – one example is SpaceX, whose IPO in June 2026 briefly made Musk a trillionaire and whose market capitalization exceeds 2 trillion USD.
AWS published a guide comparing three vector stores for Amazon Bedrock Knowledge Bases – OpenSearch, Aurora PostgreSQL with pgvector, and S3 Vectors – in terms of performance, costs, and suitability for specific RAG tasks.
Wood Mackenzie built the shared platform APEX on Amazon Bedrock AgentCore to standardize AI agent infrastructure (identity, security, observability, state management). According to the company, 88 % of internal agent prototypes never reach production; the platform is used by the applications Woody, Lens AI…
German insurance broker MRH Trowe made self-service AI agents available to 400 employees using Strands Agents, Amazon Bedrock AgentCore and LibreChat, according to the AWS blog, with costs of around 14 USD per employee per month and data in the Frankfurt region.
AWS published a guide to layered authorization for MCP tools in Amazon Quick: four gates (MFA, geographic restrictions, RBAC, tool permissions) applied to JWT claims from Microsoft Entra ID, connected through Amazon Bedrock AgentCore Gateway.
AWS described a pipeline on Amazon SageMaker AI that uses the diffusion model Qwen-Image-Edit-2509 to insert synthetic figures into real images of machinery and automatically labels them using Amazon Rekognition. According to AWS, this improved the detection of people near machinery (mAP50) by up to 160 % without manual annotation.
Heidy Khlaaf from AI Now Institute argues that existential risks from AI must be falsifiable; otherwise, they resemble religious arguments. Using nuclear weapons as an example, she shows that air-gapped systems and physical security (see Stuxnet) limit the realistic possibilities for AI to take control of them.
The assistants Instinct (Instinct Concierge) and Meta Muse can now call businesses on behalf of users – reservations, cancellations, complaints. Both features are in early access; meanwhile, Instinct is in funding talks at a valuation of 10 billion dollars, and Muse reports 730 000 downloads in its first days.
A journalist from the website 404 Media demonstrated that an AI-generated song can be uploaded through the distributor Distrokid to a verified Spotify profile of an existing band without consent – neither Distrokid nor the streaming platforms verify identity in any way.
Eric Provencher, a developer of the Codex tool at OpenAI, warns that agent swarms with many parallel sub-agents needlessly burn tokens without improving quality. He cites an example of refactoring a Python file for 20 000 dollars using 1 393 agents, which a single agent could have handled for a fraction of the price.
Weekly token consumption on the OpenRouter platform has increased by more than 25 000 percent since January 2025, from 0.5 to 126.2 trillion. According to The Decoder, this is not driven by corresponding growth in AI usage, but primarily by thinking tokens from reasoning models and unoptimized agentic systems.
According to the author of a chapter in the book Perspectives on Machine Consciousness, debate about the consciousness of future AI systems may become as divisive as abortion or climate change. A survey of 1202 participants in the USA: 45 % consider AI consciousness impossible, 23 % consider it possible, and experts also disagree.
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
