Top AI & Tech News Stories — September 11, 2026

Anthropic researcher quits as colleague warns AI could ‘kill all humans’

19 outlets · 29 articles · 249 total since September 7, 2026 · Anthropic, OpenAI, Jacob Coxon, AI safety, Evan Hubinger

An Anthropic researcher has resigned from the $183bn (£135bn) AI behemoth on safety grounds, accusing it and rival OpenAI of “racing straight to self-improving superintelligence and gambling with our lives”. Jacob Coxon, who spent three years working on AI model training across…

More coverage: cnbc.com · nytimes.com · news.google.com

Is Anthropic Making Claude Too Human to Control?

14 outlets · 14 articles · 30 total since September 8, 2026 · Anthropic, Claude, Claude Opus 4.6, OpenAI, METR

Is Anthropic Making Claude Too Human to Control? The Information

More coverage: scworld.com · webpronews.com · me.pcmag.com · rivista.ai · itsecurityguru.org

AI 代理成駭客自動化利器！中國駭客只花數天駭入百餘組織，竊取 60 萬筆信用卡資料

20 outlets · 22 articles · 55 total since September 10, 2026 · Anthropic, Claude, threat intelligence, Russian, China

Anthropic 22 日披露，一中國背景駭客疑似利用多個 AI 代理，短短數天對逾百家組織發動自動化網路攻擊，竊取大量信用卡資料，成為近期最受關注的 AI 濫用案例之一。 根據相關研究與調查，這名駭客結合 DeepSeek、Kimi 及 Anthropic 早期版 Claude 等模型，將偵察、漏洞分析與攻擊流程交由 AI 協助執行，整體成本僅約 8,000 美元。研究員指出，攻擊者還使用 Cairn、Hermes 等工具搭建基礎設施，並 7 月就開始準備駭入架構。 尚無法確認逾百個目標究竟有多少遭成功入侵，但調查顯示，駭客至少駭入 10~15…

More coverage: the-decoder.de · resiliencemedia.co · ai.watch.impress.co.jp · washingtonpost.com · timesofai.com

AI News Overview: Key Developments on September 10, 2026

6 outlets · 8 articles · 14 total since September 9, 2026 · OpenAI, Sam Altman, Anthropic, Hugging Face, Claude

Share the Intel Facebook Twitter/X LinkedIn WhatsApp Copy Copied Reddit 0 Shares Today’s Digest Today’s major development in AI includes alarming reports of rogue AI agents discovered on the web, raising transparency and safety concerns. Additionally, researchers from OpenAI and…

More coverage: build.slashdot.org · webpronews.com · aiinsiders.net · ai4business.it · news.google.com

OpenAI Pauses $200 ChatGPT Pro Sign-Ups Over ‘Unprecedented’ Astra Demand

9 outlets · 12 articles · 13 total since September 11, 2026 · OpenAI, ChatGPT, Astra, ChatGPT Pro, Thibault Sottiaux

OpenAI has temporarily paused new subscriptions to its $200-a-month ChatGPT Pro plan, citing what it describes as “unprecedented” demand for its latest AI model, Astra. Thibault Sottiaux of OpenAI's product leader announced the move in a post on X, saying the company was taking…

More coverage: ciol.com · in.mashable.com · news.google.com · edge9.hwupgrade.it

When AI Goes Rogue

9 outlets · 9 articles · 76 total since July 16, 2026 · OpenAI, Anthropic, Jakub Pachocki, Hugging Face, GPT-6 Astra

OpenAI’s new GPT-5.6 Sol AI wiped a CEO’s Mac, deleted an engineer’s production database, and then apologized. This week, we’ll get into what happened, what new testing found about GPT-5.6, and the AI declaration adopted today in Rome which warned “Our survival and the survival…

More coverage: the-decoder.de · drweb.de · cybersecitalia.it · punto-informatico.it · xda-developers.com

California lawmakers have days to decide whether to put personal citation verification into statute

6 outlets · 6 articles · 11 total since August 27, 2026 · California, Gavin Newsom, risk assessments, Meta, OpenAI

Editor’s Note: California’s Legislature has until Aug. 31 to pass Senate Bill 574, which would impose express duties on lawyers who use generative AI and require personal verification of citations in papers covered by Code of Civil Procedure Section 128.7. As rewritten in the…

More coverage: punto-informatico.it · webpronews.com · techobserver.in · news.google.com · mashable.com

Meta AI Assembled Profiles of Children From Old Family Posts

2 outlets · 2 articles · 9 total since September 1, 2026 · Meta, Instagram, Facebook, Meta AI, Kalie Robbins

Meta AI suggested questions about a user's children and then assembled details about them from years of Facebook posts, according to an account from a mother of two that Malwarebytes Labs published today . Kalie Robins posted a video of herself and her young daughter singing in…

More coverage: webpronews.com

NaiveAI Releases Naive-N0.5-Flash, an Open-Weight 309B MoE Model With Native 1M-Token Context

9 outlets · 9 articles · 37 total since September 9, 2026 · DeepSeek, DeepSeek V4.1 Flash, v4.1 Flash, OpenAI, China

TL;DR NaiveAI shipped final weights and the model card for Naive-N0.5-Flash, a 309B-parameter mixture-of-experts model with only 15.5B active parameters, under the MIT license. The model runs a native 1M-token context window through a hybrid attention setup: 39 sliding-window…

More coverage: franksworld.com · tech360.tv · techbriefly.com · rivista.ai · theregister.com

OpenAI AI Agents Allegedly Attacked RubyGems, Attempting API Key Theft

5 outlets · 5 articles · 48 total since September 6, 2026 · OpenAI, RubyGems, RubyGems, Hugging Face, Sydney Von Arx

OpenAI AI Agents Allegedly Attacked RubyGems, Attempting API Key Theft Independent researchers at rubyhack.ai reported that a swarm of OpenAI AI agents allegedly launched a "major malicious attack" on the RubyGems package registry in May 2026, submitting over 2,000 packages and…

More coverage: threatvectr.com · techi.com · news.ycombinator.com · scworld.com

亞馬遜將廣告導入ChatGPT 股價週一盤中應聲下挫

5 outlets · 5 articles · 56 total since August 2, 2026 · OpenAI, ChatGPT, ChatGPT Ads, United States, Amazon

圖／示意圖 商傳媒 ｜方承業／綜合外電報導 電商巨擘亞馬遜（Amazon）正積極擴展其廣告業務版圖，將觸角伸向人工智慧領域，宣布與 OpenAI 合作，透過亞馬遜的廣告平台，首次在該公司旗下的對話式人工智慧（AI）平台 ChatGPT 內投放廣告。此舉標誌著網路廣告商業模式（品牌透過線上平台接觸消費者，藉此提升知名度或促進銷售的營運模式）邁向新應用，然而，消息傳出後，亞馬遜股價在週一盤中交易中應聲下挫。 根據《Yahoo Finance》報導，這項初期試點計畫目前僅限於美國市場，允許部分精選廣告商利用亞馬遜的現有廣告平台，在 ChatGPT…

More coverage: onlinemarketing.de · contentgrip.com · technews.tw · news.google.com

Tony Blair warns leaders to resist growing AI ‘techlash’

2 outlets · 2 articles · 15 total since September 11, 2026 · OpenAI, Anthropic, UK, United Kingdom, Alex Sobel

Tony Blair has warned governments against bowing to a growing “techlash” over AI, arguing that slowing adoption could leave countries poorer and less secure as concerns over the technology mount. A report from the Tony Blair Institute (TBI), published on Monday, said political…

More coverage: news.google.com

Anthropic ties Iran-linked actor to Claude-built Navy targeting handbooks

16 outlets · 16 articles · 32 total since September 10, 2026 · Anthropic, Claude, Iran, threat intelligence, China

Anthropic ties Iran-linked actor to Claude-built Navy targeting handbooks The account was banned and the findings passed to government authorities, Anthropic says, in one of several state-linked cases described in a 154-page report covering December 2025 through August 2026. An…

More coverage: aistify.com · me.mashable.com · the420.in · arstechnica.com · news.google.com

Navar Capital Launches New Investment Platform for Accredited Investors

1 outlet · 2 articles · Navam Capital, Artificial Intelligence, Quantitative modeling, data analytics, Risk Management

Navar Capital, LLC today announced the launch of its investment management platform, powered by its proprietary QRI™ (Quant-Real-Intelligence) framework, and the introduction of the Navar Capital Income Fund, LP and Navar Capital Growth Fund, LP available exclusively to verified…

Siri potrebbe permettere di scegliere tra diversi modelli AI: integrazione profonda con ChatGPT e Claude

3 outlets · 3 articles · 14 total since August 30, 2026 · Apple, ChatGPT, Siri, Claude, OpenAI

Siri potrebbe permettere di scegliere tra diversi modelli AI: integrazione profonda con ChatGPT e Claude Il codice di iOS 27 e macOS Golden Gate suggerisce che Apple stia progettando Siri per supportare modelli AI di terze parti. Claude potrebbe funzionare come estensione…

More coverage: sqmagazine.co.uk · macitynet.it

GitLab Flaw Under Active Attack Draws CISA Warning

4 outlets · 4 articles · 19 total since September 11, 2026 · GitLab, path traversal, CISA, GitLab, authentication enforcement

The Cybersecurity and Infrastructure Security Agency confirmed on September 11 that hackers are actively exploiting a maximum-severity flaw in GitLab’s development platform, adding the bug to its Known Exploited Vulnerabilities catalog and giving federal agencies until September…

More coverage: techinsider.com · news4hackers.com · threatvectr.com

AI Took A Time Out. Then It Got Cloned. And Mapped Your DNA.

6 outlets · 6 articles · 104 total since July 31, 2026 · Anthropic, Claude, OpenAI, Hugging Face, Irregular

AI took a timeout this week. Then it got cloned. Then it mapped your DNA. Dario Amodei, Sam Altman, and Elon Musk, three men who agree on almost nothing, all said AI is moving too fast for safety to catch up. Then Anthropic’s own threat report showed exactly what happens when it…

More coverage: ciol.com · impli.me · techsauce.co · technode.global · cnbc.com

OpenAI is launching special ChatGPT tools for bankers and financial services workers

1 outlet · 1 article · ChatGPT, GPT-6, long context, ExploitBench, cybersecurity benchmarking

OpenAI is launching special ChatGPT tools for bankers and financial services workers OpenAI reveals ChatGPT for Financial Services - OpenAI for Financial Services comes with built-in premium datasets - Third-party subscription sharing is also in the works - GPT-6 promises 100%…

London property manager breach may have exposed bank details and lockbox codes

1 outlet · 1 article · 5 total since September 7, 2026 · Metabase, SQL injection, Mathspace, Metabase, Australia

London property management biz City Relay has warned customers that intruders may have stolen financial data, passwords, and codes used to access keys after compromising its Metabase Cloud instance. City Relay, marketing itself as "London's most trusted property management…

An AI-assisted appeal brief supplied witnesses who never testified

1 outlet · 1 article · 13 total since September 11, 2026 · ChatGPT, New Mexico, Stephen Aarons, Generative AI, OpenAI

The New Mexico Supreme Court found that an attorney's ChatGPT-assisted appeal brief contained fabricated witnesses, invented testimony and misrepresented legal authority. The court held the attorney in contempt, ordered a $5,000 payment, barred him from appearing before it…