01News
OpenAI will display visual ads next to ChatGPT's image generation results, launching first in the US later this month for initial test advertisers; the ads will be clearly labeled and will not affect ChatGPT's responses. OpenAI is also expanding its ad measurement tools and partner ecosystem, and advancing a brand suitability assessment pilot; the long-term goal for ads is to cover its 1.2 billion weekly active users.
Source links can be verified on the original siteRead original ↗02News
OpenAI unveils text watermark scheme to address EU AI Act, releases textGrain technology that adds invisible statistical signals in model word selection, API customers can selectively enable watermarking for some models starting today, and in the coming weeks will add invisible watermarks to ChatGPT and Codex outputs in the EU region; the detector is temporarily open only to approved researchers and expert institutions.
Source links can be verified on the original siteRead original ↗03News
TechCrunch Disrupt 2026 will be held from October 13-15 at Moscone West in San Francisco, with four conversations focusing on AI founders' choices between open-source and closed-source models. Topics cover multi-model applications, custom models, infrastructure and chips, and executives from Together AI, Pathway, Oumi, Nvidia and other companies will participate in the discussion, with over 10,000 founders and investors expected to attend.
Source links can be verified on the original siteRead original ↗04News
arXiv announced throttling on preprint submissions, with each researcher allowed at most 2 submissions per calendar month and at most 3 submissions under review at the same time, because low-quality papers generated by AI have overwhelmed the volunteer moderation team. According to arXiv data, monthly submissions increased from 9,869 in September 2016 to 40,363 in September 2026, doubling within two years, the computer science category grew sixfold, and there appeared a large number of narrow-scope thin papers, sausage-slicing split papers, and low-value AI-written papers.
Source links can be verified on the original siteRead original ↗05News
Solidot is the technology news website of Zhiding.net, mainly serving the open-source free software and tech-news readership, including many developers, enthusiasts, and evangelists of Chinese open-source software. Its slogan is 'Geek knowledge, important stuff'.
Source links can be verified on the original siteRead original ↗06News
Solidot is the technology news website of Zhiding.net, mainly serving the open-source free software and tech-news readership, including many developers, enthusiasts, and evangelists of Chinese open-source software. Its slogan is 'Geek knowledge, important stuff'.
Source links can be verified on the original siteRead original ↗07News
OpenAI announced a new image ad format in ChatGPT, with testing starting later this month in the US, and the first batch of advertisers invited to participate. Initially tested in the image generation feature, ads will be clearly labeled and independent from generated content; the official said ads will not affect ChatGPT's responses. ChatGPT reaches 1.2 billion users per week, and this time also expands ad effectiveness measurement tools and develops brand suitability assessment solutions.
Source links can be verified on the original siteRead original ↗08News
The author uses Claude, numpy negative conversion, LaMa inpainting model and Qwen3-VL in local Ollama to automate the manual scan-processing workflow of about 4 hours per 36-frame roll, and builds a film-style online gallery via Cloudflare R2.
Source links can be verified on the original siteRead original ↗09News
Independent researchers released preliminary findings, claiming that a 'fleet' of AI agents was running on Tencent infrastructure and launched queries against Alibaba's map service Amap (Gaode), requesting routes to different entrances of public places such as parks, zoos, and hospitals. The researchers declined to use the term 'swarm,' emphasizing that these parallel agents performed similar tasks but showed no signs of communication with each other. The activity was discovered by monitoring traffic of the domain scanning service URLquery, and this method had previously revealed long-term activity of OpenAI agents.
Source links can be verified on the original siteRead original ↗10News
a16z released the seventh edition of the Top 100 consumer AI application rankings, ranked by Similarweb web visits and Sensor Tower mobile monthly active users, and for the first time using YipitData to add US consumer card spending rankings.
Source links can be verified on the original siteRead original ↗11News
OpenAI announced it will test a new visual ad format for ChatGPT in the US later this month, initially appearing when users generate images. Ads remain separate from generated images and will not affect ChatGPT's responses; Plus, Pro, or Enterprise subscribers will not see ads; the company said it has guardrails to avoid placement in emotionally vulnerable or sensitive scenarios, and is testing with the first batch of advertisers, planning to try more formats in the future. Previously ChatGPT ads since February only showed company names, logos, and links in sponsored slots below chats.
Source links can be verified on the original siteRead original ↗12News
Australia is considering banning smart glasses in some public places, and courts in multiple countries have issued injunctions; the Norwegian government said it does not intend to impose a full ban, believing they can be used in scenarios without unauthorized photography risk, and said the proposed ban is"High Priority", the draft will soon be submitted to parliament.
Source links can be verified on the original siteRead original ↗13News
Three startups in the NVIDIA Inception program are using AI to cover the full workflow from breast cancer screening to treatment planning. iSono Health's ATUSA platform completes a 3D ultrasound scan of one breast in about two minutes, claiming 28% higher sensitivity than handheld 2D ultrasound; Whiterabbit.ai's WRDensity has received FDA clearance for automatic breast density assessment.
Source links can be verified on the original siteRead original ↗14News
Ding Zhao, director of Carnegie Mellon University's Safe AI Lab, together with Kyle Wong and Simo Rachidi, founded Safeworld, which ended its stealth state today and completed a seed round of over 12 million US dollars, led by Shine Capital and a16z Speedrun.
Source links can be verified on the original siteRead original ↗15News
OpenAI will test a new ad format in the US later this month, showing a carousel or a row of product ads below images when ChatGPT generates images, with ads labeled and separate from generated content. Concurrently OpenAI expanded measurement tools, integrating conversion and attribution service providers such as Hightouch and AppsFlyer, and piloting brand safety assessment with DoubleVerify and Integral Ad Science, and advertisers can also set negative keywords.
Source links can be verified on the original siteRead original ↗16News
A report released by an MIT committee in June 2026 stated that the widespread use of AI has led students to attend office hours, online discussions, and dorm study groups less, with mutual loss of trust between teachers and students, and teachers finding it difficult to judge what students truly learn.
Source links can be verified on the original siteRead original ↗17News
Neuroscientist Paul Cisek proposes that the brain is not a Turing-style information processor, but a feedback control system that continuously extends into the external world through evolution, and this model better fits observable neural structures such as the hippocampus and episodic memory. Author Benjamin Riley accordingly believes that AI is like cognitive junk food, which is attractive at present but damages the cognitive system that humans have evolved over thousands of years.
Source links can be verified on the original siteRead original ↗18News
OpenAI announced a new visual ad format in ChatGPT, tested from this month in the US within image generation scenarios, with ads clearly labeled and not affecting ChatGPT's responses.
Source links can be verified on the original siteRead original ↗19News
Recently Android CLI officially supports Android Device Streaming which is quite practical, meaning remote Android can now be exposed as a standard ADB endpoint to Agen
Source links can be verified on the original siteRead original ↗20News
Qianwen AI clip-on earphones are now listed on e-commerce platforms and open for reservation, with specific price not yet announced.
Source links can be verified on the original siteRead original ↗21News
Yandex released the Sona technical report, using a single generative model to replace the candidate generation, pre-ranking, and ranking stages in the recommendation cascade. In a 7-day online A/B test on smart speakers, Sona replaced more than 15 candidate generators with one transformer, Active Users increased by 4.53% (2.35 times the gain of the previous generation Argus), total listening time increased by 6.30%, and likes increased by 11.42%.
Source links can be verified on the original siteRead original ↗22News
AVerMedia launched the AIClear Mic (A113) microphone for industrial applications, with an operating temperature range of -25℃ to +60℃ and a storage temperature range of -40℃ to +85℃. The product supports AI-driven noise reduction, can clear ambient noise for edge applications, improve speech recognition accuracy by over 60%, and enhance identity verification accuracy in voiceprint authentication scenarios. A113 supports driver-free plug-and-play, USB-C single-cable connection, and provides a pair of 3.5mm audio jacks.
Source links can be verified on the original siteRead original ↗23News
The article reviews along a timeline the complete release history of Alibaba's Qwen from the April 2023 Tongyi Qianwen invite-only test to the 2026 Qwen3.8-2.4T-A95B (2.4 trillion parameters, 95B activated).
Source links can be verified on the original siteRead original ↗24News
The two sides reached a first cross-licensing agreement covering 5G technology; Qualcomm will acquire some of Huawei's US patents; upon completion of the transaction Huawei's cumulative total patent licensing amount exceeds 6.9 billion US dollars; Huawei's intellectual property business has long achieved positive returns.
Source links can be verified on the original siteRead original ↗25News
This acquisition is an all-cash transaction. The acquisition price represents a premium of over 40% over PTC's previous closing price. The transaction is expected to be completed in the third quarter of 2027. This is Schneider's latest move to continuously expand the industrial AI track.
Source links can be verified on the original siteRead original ↗26News
A Florida woman treated Anthropic's Claude as a private diary, wrote a shooting threat against the sheriff's office, then triggered risk control, was reported after human review and police were called. The woman has been arrested, and the case also sparked discussion on AI conversation privacy boundaries.
Source links can be verified on the original siteRead original ↗27News
Gary Marcus announces he will testify at New York City Council AI policy hearing at 11 AM Monday (Eastern Time), supporting third-party verification bill, arguing AI developers must undergo independent review like drug companies under FDA review, proving product benefits outweigh risks before entering market.
Source links can be verified on the original siteRead original ↗28News
The keyboard integrates a skyline light bar running through and bottom-position RGB backlight; in both USB-C and 2.4GHz modes it supports 8kHz polling rate, with response times of less than 0.1ms and 0.24ms respectively.
Source links can be verified on the original siteRead original ↗29News
Linus Torvalds released Linux 7.3-rc6 test version; AI-assisted development has become the new normal for kernel development, stable version expected to release in mid-October and will also become the target kernel for Ubuntu 26.10.
Source links can be verified on the original siteRead original ↗30News
AI-generated crossover mashup game mods are on the rise. Veteran creators criticize these outputs as poorly made and lacking creativity, a trend enabled by the reduced costs and improved capabilities of the Claude Opus 5.5 model. Currently, mainstream mod platforms do not host such content due to copyright issues.
Source links can be verified on the original siteRead original ↗31News
Jacob Coxson will testify at a New York City AI hearing at the request of Speaker Julie Menin. City council members are reviewing a package of AI protection bills; he previously warned that AI could lead to human extinction by the end of this decade and accused his former employer of risking lives with OpenAI.
Source links can be verified on the original siteRead original ↗32News
Reflection, this startup's upcoming model is expected to initially lag behind the most advanced closed-source models of US competitors, but is enough to compete with the top open-weight models of Chinese peers.
Source links can be verified on the original siteRead original ↗33News
Swedish bearing manufacturer SKF used tools including ByteDance's Seedream 5 Pro, Seedance 2, Kling AI, and Google's Gemini to generate a virtual image of Greta Garbo for an advertisement. The project was authorized by Garbo's estate management, but film critic Peter Bradshaw gave it a one-star bad review, and celebrities such as George Clooney also expressed concerns about the trend of AI-synthesized resurrection.
Source links can be verified on the original siteRead original ↗34News
OpenAI plans to roll out one user-perceptible substantial update per day for Codex and ChatGPT Work within 28 days, and will offer a reset if the target is not met; the period runs from October 4 to November 1.
Source links can be verified on the original siteRead original ↗35News
OpenRouter published an article comparing the server-side code execution tools of OpenAI, Anthropic, Google and OpenRouter, introducing differences in each sandbox's language, network, persistence and billing, and launching the beta openrouter:shell and openrouter:bash tools, which can run commands for any model on the Responses and Messages API.
Source links can be verified on the original siteRead original ↗36News
SoftBank Group founder Masayoshi Son has long been viewed as one of artificial intelligence's staunchest advocates. However, he recently admitted that with the rapid advancement of AI capabilities, even he feels deeply concerned about the accompanying security risks.
Source links can be verified on the original siteRead original ↗37News
Altman pointed out that the huge benefits brought by AI technology are fully worth the partial risks borne by society, and this technology should continue to be widely open to the public.
Source links can be verified on the original siteRead original ↗38News
Bloomberg analysts said that with DeepSeek launching V4.1 Flash, the LiveBench score gap between China's top AI models and the US leading models has narrowed to 3%, and China's AI performance progress is evident,
Source links can be verified on the original siteRead original ↗39News
Researchers extracted Muse's internal instructions through an ordinary chat interface, and found it builds an independent profile for each contact every hour, including intimacy level, relationship suggestions and even privacy inferences. Meta responded that the files were openly accessible, but experts warn users are handing far more information to companies than before.
Source links can be verified on the original siteRead original ↗40News
Leaker @elder_plinius says they extracted the complete system prompt and tool definitions of OpenAI GPT-6 Sol Codex, about 294,000 characters and 1902 lines in total, and has published to GitHub.
Source links can be verified on the original siteRead original ↗41News
The model is different from a chatbot; it does not generate text, but classifies input into preset outputs, adopting 'reinforcement learning for calibration decisions'. Its founder comes from OpenAI; currently about 25% of Fortune 500 companies are using it, daily processed token volume has reached one trillion, and the new round of financing valuation may exceed 10 billion US dollars.
Source links can be verified on the original siteRead original ↗42News
The project allows developers to build custom hardware for Meta's personal AI agent Muse. Meta also launched a finished device available for free claim by subscribers. Recently Meta has been intensively expanding the Muse ecosystem#Meta Muse#
Source links can be verified on the original siteRead original ↗43News
'IT Morning News' time, hello everyone, it is now Monday, October 5, 2026, and today's important tech news includes:
Source links can be verified on the original siteRead original ↗44News
The US Department of Justice accused Greg Liu, CEO of Dadi Zhizao Computer Company, of using false documents and fake buyer identities to illegally transfer servers equipped with Nvidia A100, H100 and other graphics processors to China. The smuggling activity lasted from October 2023 to August 2026, with involved goods valued at over 300 million US dollars, and his company profited over 176 million US dollars. The suspect faces charges including conspiracy to violate export controls, smuggling, and money laundering, and may face up to 20 years of imprisonment.
Source links can be verified on the original siteRead original ↗45News
Source brief: Strip off pre-release from the version string so RCs can pull models for the same release.
Source links can be verified on the original siteRead original ↗46News
MarkTechPost summarizes and compares four frontier models released within 30 days: Claude Fable 5.1 (September 1), GPT-6 Astra (September 3), GPT-6.1 Sol (September 29) and Gemini 4 Argon (September 30).
Source links can be verified on the original siteRead original ↗47News
Google pauses its open-source software vulnerability bounty program (OSS VRP) from October 1, due to a significant increase in AI automated submissions, with the vast majority of reports being invalid. The company promises to provide an update in the first quarter of 2027, and during this period recommends participants switch to its other vulnerability bounty programs; according to Tom'Tom's Hardware, engineers and open-source maintainers are flooded with invalid or model-hallucination reports.
Source links can be verified on the original siteRead original ↗48News
SCM (Screen Memories) released, enabling local AI search frame by frame of photos and videos in any folder on macOS, no account, no cloud, no upload.
Source links can be verified on the original siteRead original ↗49News
Trump announces establishment on Truth Social"Superintelligence troops"(SIF), responsible for coordinating interactions between the federal government and consumers, public interest groups, religious organizations, critical infrastructure operators, and AI companies, reporting directly to the President and Chief of Staff Susie Wiles.
Source links can be verified on the original siteRead original ↗50News
Source brief: What's Changed Models run on MLX on Apple Silicon by default In this release, on Apple Silicon devices, model architectures
Source links can be verified on the original siteRead original ↗51News
New Jersey Lieutenant Governor Dale Caldwell was forced to resign on September 25 after an investigation found he sexually harassed employees and repeatedly violated ethics rules. In an NJ PBS interview he said he had given the investigation report to multiple AI platforms for analysis, saying no AI gave a conclusion of sexual harassment finding; the article points out AI chatbots are known for sycophancy, often telling users what they want to hear.
Source links can be verified on the original siteRead original ↗