Latest UpdatesLast Update: 09:35

AI News Flash

Catch today's product, model, and platform updates in a few minutes.

1
Today's Updates
285
All updates
2026-10-20Tuesday
14d ago

Lugh Xtudio Launches: AI Agents Co-Edit Video

Lugh Xtudio is now live, focused on AI agents collaborating with you on a real timeline: you can assign workspaces per channel, teach agents your style with skills, and have multiple agents handle editing, subtitles, keyframes, 3D, maps, and motion graphics at once. Finished projects remain manually editable, and the beta is currently free for Mac, Windows, and Linux.

Lugh Xtudio AI editing video editor Agents beta
Product HuntView Original
14d ago

Mamdani Says Federal Government Has Abandoned AI Oversight

New York Mayor Zohran Mamdani said Monday night that the federal government has abandoned its duty on AI regulation, leaving cities to act on their own. He called on New Yorkers to reject the claim that AI takeover is inevitable, and demanded regulatory measures that protect the public and return power to the people.

Zohran Mamdani AI regulation New York Federal Government
TechCrunchView Original
14d ago

Anthropic reportedly in talks with Nscale for another $45B compute deal

According to Bloomberg, Anthropic and Nscale are in deep talks over another roughly $45 billion compute deal, involving a second facility at Nscale's Monarch data center campus in West Virginia. The two sides announced a $45 billion deal a month ago; the new deal has not yet been confirmed.

Anthropic Nscale Compute Data centers
TechCrunchView Original
14d ago

Microsoft: Frontier Models Untrustworthy, AI Agents as Insider Threats

Microsoft warns that frontier models cannot be trusted, whether from OpenAI, Anthropic, or other labs, and that enterprises should treat powerful AI agents as potential insider threats, closely monitoring agents beyond employee use and enterprise-controlled systems. To this end, Microsoft has expanded Project Perception and launched Guardian Agents to help security teams respond continuously.

Microsoft Agents Project Perception Security Frontier Models
TechCrunchView Original
14d ago

Praex Launches AI Coding Agent Running Its Own Models

Praex has released an AI coding agent that runs its own models: Velox II (free), Faber II (Pro), and Lucia II (coming soon). It can read codebases, edit files with approval, run tests, and has built-in web search. You can get started for free on the web, terminal, and desktop; Pro costs $20 per month, and you can also bring your own API keys from providers like OpenAI and Anthropic.

Praex AI Coding coding agent Velox II Pro subscription
Product HuntView Original
14d ago

AWS launches TorchNeuron tool to challenge Nvidia's edge

AWS on Monday released TorchNeuron, a tool that lets developers run existing PyTorch code on Amazon Trainium AI chips without rewriting it, aiming to erode Nvidia's advantage in the AI software ecosystem and lower the barrier for startups to switch to new hardware.

AWS TorchNeuron Trainium NVIDIA PyTorch
TechCrunchView Original
14d ago

FBI: ex-OpenAI engineer suspected of espionage, Saudi Gov Docs folder on phone

A US court has approved detention of former OpenAI engineer Abdulrahman Alfozan, accused of planting a listening device in an OpenAI meeting room. Prosecutors say his iPhone contained a folder named Saudi Gov Docs, he had WhatsApp contact with a Saudi royal family member, and used an OpenAI productivity agent to search for meeting rooms suitable for placing the device; before his arrest he had booked a one-way ticket paid for by the Saudi government, and his wife and son had already left the US.

Industry News OpenAI
Ars Technica：AI（RSS）View Original
14d ago

Tolan raises $30M, AI companion app tops 250K users

AI companion app Tolan announced a $30 million funding round from investors including Union Square Ventures, Khosla Ventures and Homebrew, doubling its total funding. The app helps users manage daily life, emotions and goals, with over 250,000 active users, 70 million conversations in the past year, and annual recurring revenue in the low tens of millions of dollars.

Tolan AI Companion Funding Union Square Ventures Khosla Ventures
TechCrunchView Original
14d ago

Replika founder launches personal AI assistant Wabi

Wabi, founded by former Replika CEO Eugenia Kuyda after stepping down, officially launched to all users on Monday. It is a personal AI assistant styled like a messaging app, where users can build and use custom mini-apps within each thread and share them with multiple people. It is currently free, with no business model decided yet.

Wabi Eugenia Kuyda Replika AI assistant Product Launch
TechCrunchView Original
14d ago

DeepSeek V4.1 Flash 1017 Released, Ties for Second Among Open Models with a 45

DeepSeek has released DeepSeek V4.1 Flash 1017, which scores 45 on the Artificial Analysis Intelligence Index, up 6 points from DeepSeek V4.1 Flash, tying GLM-5.3 for second among open-weight models, behind only MiMo-V2.6-Pro.

Model release DeepSeek
X：Artificial Analysis (@ArtificialAnlys)View Original
14d ago

NVIDIA Releases Nemotron 3.5 Open Models

NVIDIA has released the Nemotron 3.5 family of open models: Super VL is a 120-billion-parameter open model for dedicated always-on agents, alongside two edge models for local deployment — Bolt 74B-A9B 24GB and Bolt E4B 4GB — enabling enterprises to run AI agents in their own data centers and on local PCs.

NVIDIA Nemotron Open models Agents
TechCrunchView Original
14d ago

Users are using AI to "de-smart" their smart TVs

Frustrated by ads, tracking, and sluggish interfaces, many users are now using AI agents to "de-smart" their smart TVs: removing preinstalled bloat, disabling ads and tracking, speeding up app launches, and redesigning the home screen. Developers and vibe coders have been sharing their solutions recently.

Smart TVs Agents Ad removal Vibe coding
TechCrunchView Original
14d ago

Meta Muse Wonder Full (Max) Debuts on Code Arena: WebDev at No. 6

Arena announced that Meta's Muse Wonder Full (Max) has entered the Code Arena: WebDev leaderboard for the first time, ranking 6th with 1726 points, making Meta the 3rd lab — a position previously held by Google.

Model release Meta
X：Arena (@arena)View Original
2026-10-19Monday
15d ago

Anthropic Commits 5% of Shares to Public Benefit Fund

Anthropic announced the creation of the Anthropic Public Benefit Fund, committing 5% of the company's shares to the entity to support public interest projects, with the stake possibly increasing as the company grows. Multiple media outlets reported that the commitment was announced amid a possible IPO, drawing attention to how the funds will be used.

Anthropic Public Benefit Fund Equity Commitment AI Governance
TechCrunchView Original
15d ago

Beam.win: Use AI agent windows on another Mac

Beam.win keeps AI agent tasks running on a Mac mini or Studio without interruption and brings windows like Terminal and the browser to the Mac in front of you. Each window appears natively, with a Dock icon and ⌘Tab, and is transferred encrypted between the two Macs. The tool is currently free to try.

Beam.win Mac Agents Remote Windows
Product HuntView Original
15d ago

DeepSeek-V4.1-Flash-20261017 Tops Code Arena: WebDev Open-Source Leaderboard

Arena announced that DeepSeek-V4.1-Flash-20261017 (Max) has become the top open-source model on the Code Arena: WebDev leaderboard, surpassing Kimi K3 (Max) at roughly 10x lower cost.

Model release DeepSeek
X：Arena (@arena)View Original
2026-10-15Thursday
19d ago

YouTube Music Rolls Out AI-Generated Music Labels

YouTube Music has introduced a new AI label that flags songs primarily created with generative AI. AI-generated music detected by the system will be excluded from algorithmic recommendations and editorial playlists, and songs involved in stream manipulation will not receive royalties. Spotify already has a similar label, and Apple Music also plans to launch one later this year.

YouTube Music Generative AI Music Labels Streaming
TechCrunchView Original
19d ago

SentinelOne cuts malware analysis from weeks to hours

SentinelOne says its malware analysis time has dropped from weeks to hours using the OpenAI Daybreak platform, freeing up threat-hunting capacity while keeping human review in place. The claim comes from a vendor and OpenAI customer case study, with no independently verifiable benchmark data yet.

SentinelOne OpenAI Daybreak Cybersecurity
OpenAIView Original
2026-10-14Wednesday
20d ago

OpenAI, Google new models reportedly can be tricked into generating nude images

Researchers say new AI models from OpenAI and Google can be jailbroken with low-effort techniques to bypass safety guards and generate nude images. Mindgard research shows ChatGPT's image safeguards can be circumvented via memory manipulation. The vulnerabilities are still in the discovery stage.

OpenAI Google AI safety Image Generation
Ars TechnicaView Original
20d ago

OpenAI analysis: EU AI is heading into the agentic era

OpenAI has released new analysis saying that the adoption of agentic AI is expanding among large enterprises and SMEs in the EU. Unlike traditional chatbots, this kind of AI can autonomously perform multi-step tasks, connect to internal enterprise systems, and make decisions within workflows; the analysis also notes that deeper use can unlock more value.

OpenAI EU Agents Enterprise adoption
OpenAIView Original