Skip to content

AI Flash · live feed

What is happening in AI right now

A timeline of everything that has passed through the radar — newest first, with an evidence status for each item.

6 new news items today 1139 across the entire feed ↗ details, original sources and our own analyses

Continuously updated

Latest items

1139 items in the selected feed

How to read evidence statuses ↓

Tuesday, July 28

12 items
Tools and apps only one source so far

AI Mode in Google Search: planning offline activities

Google introduces AI Mode features in Search for planning offline activities — finding local classes, choosing equipment, buying tickets to events, and planning dinners. The system relies on personalization through integration with Google Calendar and other apps.

Google AI blog (The Keyword — AI) Original source ↗
AI agents only one source so far

AI enters private markets

Startup F2 uses generative AI from Anthropic, Google and OpenAI with model context protocols (MCP) to automate analysts' work in private equity and private credit. The aim is to standardize unstructured data and accelerate processing of investment deals that traditionally require manual work.

Security only one source so far

Anthropic CEO Amodei highlights the risks of open weights in AI models

In a blog post, Dario Amodei maintains that Anthropic has never called for a ban on open weights, but sees them as a significant risk involving authoritarian states and misuse for cyber or biological attacks. He calls for stricter controls on chip exports to China, action against model distillation and mandatory…

The Decoder (daily AI news) Original source ↗
Regulation and law only one source so far

A new ITU commission for global AI governance and its potential for Africa

The International Telecommunication Union (ITU) established the AI for Good Commission on 2 July 2026 with 40 members from governments and organizations. Its aim is to build trust, promote responsible innovation and increase human capacity. The commission operates alongside other global AI governance bodies; in Africa it has access to…

Tech Policy Press Original source ↗
Hardware only one source so far

The Stream smart ring: A voice interface without a phone

The Stream smart ring from Sandbar focuses on voice transcription and communication with AI assistants without requiring a phone. The device functions as a Bluetooth microphone, supports whispered commands and is expected to be released later in 2026. It competes with other AI gadgets focused on recording conversations.

Tools and apps clearly official source

GitHub Copilot for JetBrains: Improved OpenTelemetry configuration and model management

The GitHub Copilot update for JetBrains brings OpenTelemetry configuration for agent workflows, configurable token limits (maxInputToken, maxOutputToken) and support for MCP servers in Claude agent flows. Copilot CLI now supports the /rubber-duck command and displays a todo list. Enterprise users see…

GitHub Changelog (Copilot and AI features) Original source ↗
Regulation and law only one source so far

Dario Amodei (Anthropic) denies supporting a ban on open-weight models, warns of AI misuse by authoritarian governments

Anthropic chief Dario Amodei denied that the company supports a ban on open-weight models, including Chinese models. He was responding to a letter from Nvidia, Meta, Microsoft, Mistral and Hugging Face opposing restrictions. He sees the main threat in authoritarian governments, not in open weights as such.

Monday, July 27

12 items
AI agents only one source so far

A guide to choosing AI tools — from chats to agents

Ethan Mollick analyzes the shift from chat interfaces to agentic systems. ChatGPT offers Work and Codex modes, while Claude has Cowork and Code, both with access to the user's computer. Gemini lags behind in the agent category.

Simon Willison — AI tag (leading independent LLM commentator) Original source ↗
Tools and apps clearly official source

The GitHub Copilot app has its own enterprise-level access policy

The GitHub Copilot app now has a dedicated policy for managing access at enterprise and organization level, separate from the GitHub Copilot CLI policy. It enables administrators to control each client independently. The default is enabled. Changes are made through pull requests with standard audits and…

GitHub Changelog (Copilot and AI features) Original source ↗
Regulation and law only one source so far

Global asymmetry in accountability for artificial intelligence

An analytical text about the gap in global accountability for artificial intelligence. It highlights the concentration of AI production in a few companies and countries, asymmetries in impacts (particularly on the Global South), regulatory gaps, legal uncertainty and power imbalances between large technology companies and states.

Tech Policy Press Original source ↗
Business and investment only one source so far

Verizon launches AI Connect with a Google deal worth a billion

Verizon and Google reached a billion-dollar agreement to connect data centers. Verizon is launching AI Connect and building edge AI infrastructure by converting central offices into small data centers. The CEO plans further AI deals worth billions of dollars by the end of the year.

Ars Technica (AI) Original source ↗
Productivity only one source so far

Efficiency with AI is about understanding the harness, not tricks

The author argues that productivity with AI comes from a deep understanding of the existing harness rather than new tools or obscure tricks. The author recommends starting with GitHub Copilot CLI and using agent autonomy (YOLO mode), while highlighting the security risks of local execution.

Regulation and law only one source so far

Delhi High Court dismisses ANI's copyright lawsuit against OpenAI

Delhi High Court dismissed ANI's request for an injunction against OpenAI. ANI could not prove that GPT-4 and GPT-4o reproduce articles verbatim; the articles used as evidence dated from after the model's training data. The court found no economic harm and recognized language models' contribution to education.

The Decoder (daily AI news) Original source ↗
Other only one source so far

AI-powered cognitive radar and electronic warfare systems

Overview of the application of AI/ML (neural networks, fuzzy logic, genetic algorithms) in adaptive radar and electronic warfare systems for defense against mode-agile threats that defeat traditional static database systems. Description of the architecture and validation of algorithms in simulation environments.

IEEE Spectrum — Artificial Intelligence Original source ↗
Tools and apps clearly official source

GitHub extends central management of Copilot settings to the Copilot app and the cloud agent

GitHub has extended central settings management (managed-settings.json) for Copilot to the Copilot app and the cloud agent - previously, it applied only to Copilot CLI and VS Code. Administrators can thus manage plugins, marketplaces and the bypassing of approval prompts consistently across clients.

GitHub Changelog (Copilot and AI features) Details and sources →
Security only one source so far

ChatGPT refuses direct requests to imitate authors' styles

ChatGPT now refuses to generate texts that precisely imitate the styles of living authors (J.K. Rowling, Amy Tan). Instead, it offers texts drawing on their qualities but in a distinct voice. An analysis by No Latency found that it still complies with requests involving deceased authors.

Ars Technica (AI) Original source ↗
Tools and apps only one source so far

Meta launches the Meta AI assistant in private Threads messages

Meta is integrating its Meta AI chatbot into private messages on the Threads social network. Previously, AI interaction was available only publicly. Users can share posts, images, links and videos with the AI. Global rollout starts on Monday. Meta's goal: Keep users in its ecosystem and compete with tools…

TechCrunch AI Original source ↗