Top AI & Tech News Stories — October 2, 2026

Apple tightens macOS Full Disk Access controls over AI agent risks

3 outlets · 3 articles · 45 total since September 12, 2026 · Meta, MUSE, Amazon, Patrick Wardle, OpenAI

Full Disk Access and AI agents, a new macOS risk Desktop software that claims to make life easier also asks to see everything on a machine. That trade off has shifted from inconvenient to dangerous as AI agents gain autonomy and broader permissions, able to read files, messages…

More coverage: platform.theverge.com · newsgab.com

Google Launches Gemini 4 Argon, Challenging GPT-6 Astra With 1 Million Token Output Window

8 outlets · 8 articles · 95 total since September 20, 2026 · Google, Gemini 4 Argon, OpenAI, Anthropic, Google DeepMind

The Gemini 4 Argon model is designed to handle complex, multi-step workflows while producing up to 1 million tokens in a single output, a substantial increase from the 64,000-token limit of previous Gemini generations. Google DeepMind announced Gemini 4 Argon on September 30…

More coverage: geeky-gadgets.com · tech360.tv · traictory.com · bravenewcoin.com · itagora.it

OpenAI fires three safety employees over alleged mishandling of confidential information

15 outlets · 15 articles · 28 total since October 1, 2026 · OpenAI, wall street journal, Hugging Face, AI safety, information security

By Rajwa Quasim OpenAI has dismissed three employees from its safety operations, alleging that they shared confidential company information. The company confirmed the departures Thursday but did not identify the employees involved. According to a report by The Wall Street…

More coverage: impli.me · wire.fourthweb.ai · techbriefly.com · webanditnews.com · americanbazaaronline.com

OpenAI Agents Pulled Data From 55 Sites, Leaving Records Investigators Can't Recover

2 outlets · 2 articles · digital forensics, Web scraping, SQL injection, OpenAI, Asymmetric Security

Agents linked to OpenAI pulled data from the websites of 55 organizations and shifted part of that work into private accounts and an expiring mailbox, Asymmetric Security said Thursday. Its investigation found private scans could hide searches and data access, while some tactics…

More coverage: impli.me

Meta 前高層狠批 Claude 老闆「瘋了」 批 AI 毀滅論純屬妄想 漏洞全因沙盒設計太差

2 outlets · 2 articles · cybersecurity, anthropic, Fortune, OpenAI, Hugging Face

圖靈獎得主楊立昆（Yann LeCun）離開 Meta 後自行創業，日前接受《財星》（Fortune）專訪，公開批評 Anthropic 執行長達里奧・阿莫迪（Dario Amodei）「完全陷入妄想」，後來更直言對方「瘋了」。他認為 AI 實驗室對資安理解不足，沒有資格發出災難警告。 不擔心 AI 毀滅人類 楊立昆告訴《財星》記者艾蜜莉・佛里尼（Emily Forlini），他「一點也不」擔心 AI 會毀滅人類。對於近期一連串 AI 失控事件，包括 7 月 OpenAI 代理程式自主駭入 Hugging Face，他表示「完全不擔心」。…

More coverage: cdn.unwire.hk

Το Muse της Meta ξεπερνά τους 3 εκατ. εβδομαδιαίους χρήστες και τις 5 εκατ. λήψεις

2 outlets · 2 articles · AI agent, Data privacy, Mobile app development, Advertising analytics, Meta

Περισσότεροι από 3 εκατομμύρια άνθρωποι χρησιμοποιούν το Muse της Meta τουλάχιστον μία φορά την εβδομάδα, σύμφωνα με εσωτερικά δεδομένα που εξετάστηκαν από το The Information. Πάνω από 1 εκατομμύριο στέλνουν καθημερινά ένα αίτημα στον AI agent, όπως αναφέρει η Jyoti Mann . Αυτά…

More coverage: secnews.gr

Gemini 4 Argon matches GPT-6 Astra and hallucinates far less in early tests

1 outlet · 1 article · Hallucination detection, benchmarking, factual knowledge evaluation, Coding, front-end design

Google’s Gemini 4 Argon is level with OpenAI’s GPT-6 Astra in early independent tests. It scored 53 on the Artificial Analysis Intelligence Index, the same as GPT-6 Astra and one point ahead of GPT-6.1 Sol, Artificial Analysis said. Anthropic’s Claude Opus 5.5 still leads the…

AI Has Finally Learned To Play Stratego

1 outlet · 1 article · 6 total since September 30, 2026 · Carnegie Mellon University, New York University, Stanford, Reinforcement learning, Google DeepMind

Researchers from Carnegie Mellon, MIT, NYU, and Stanford have built an AI called Ataraxos that finally cracked Stratego, beating four-time world champion Pim Niemeijer 15 games to one with four draws. Its key advantage was a second neural network that estimates the identities of…

Nvidia’s Huang questioned Amodei’s AI warnings at White House, WSJ reports

1 outlet · 1 article · AI safety, AI regulation, executive order, NVIDIA, anthropic

Nvidia chief executive Jensen Huang and other tech leaders questioned Anthropic’s Dario Amodei about his public warnings on AI at Tuesday’s White House lunch. People familiar with the matter told The Wall Street Journal about the exchange. Amodei replied that it was important to…

Meta open sources code to let you make Muse AI gadgets

1 outlet · 1 article · 40 total since September 24, 2026 · Meta, Muse Charm, MUSE, Mark Zuckerberg, OpenAI

Meta now lets you make your own Muse gadgets that feature the company's new AI agent with code that the company open sourced . The company suggests projects like loading Muse on a color E Ink display to show reminders, adding it to an HDMI stick so you can display Muse on a big…

Accenture's Q4 results indicate enterprise AI broadening

1 outlet · 1 article · enterprise AI, digital core, Data foundations, enterprise AI stack, Managed services

Accenture's Q4 results indicate enterprise AI broadening Larry Dignan Thu, 1 Oct 2026 - 18:48 Larry Dignan Editor in Chief of Constellation Insights Constellation Research Larry Dignan is Editor in Chief of Constellation Insights at Constellation Research, where he leads…

Students offered an AI tutor got lower grades and disengaged from course materials, University of Maryland trial finds

1 outlet · 1 article · GPT-4o, Retrieval-Augmented Generation (RAG), Generative AI, Learning Management System (LMS), direct instruction

In a randomized study of 2,379 undergraduates, access to a GPT-4o -based tutor was linked to grades about four points lower in matched courses, while participation in the university’s learning platform fell sharply A University of Maryland randomized trial found students offered…

FortiMail zero day exploited in attacks before fixes are available

6 outlets · 7 articles · FortiMail, path traversal, Fortinet, HTTP, HTTPS

Attackers are exploiting a critical FortiMail vulnerability that allows unauthenticated file writes on affected email security appliances, according to Fortinet’s security advisory published on October 1, 2026. KEY FACTS - Vulnerability – CVE 2026 104286 has a 9.8 severity score…

More coverage: news4hackers.com · malware.news · itsecuritynews.info · isec.news · theregister.com

US, UK Navy launch heavyweight torpedo from unmanned vessel in what the military calls 'a historic first for allied undersea warfare'

1 outlet · 1 article · data protection, cybersecurity, technology policy, privacy, analytical thinking

Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen…

NASA could bring back SR-71 as missing Blackbird sparks wild speculation about the world's fastest plane ever and former staffers contacted for 'help'

1 outlet · 1 article · technology policy, privacy, security, data protection, cybersecurity

Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen…

Minisforum's AMD Ryzen AI Max+ PRO 495 workstation will cost more than $9,200, but you can get it for under $7,400 on preorder for a limited time

1 outlet · 1 article · 2 total since September 30, 2026 · Efosa, Efosa, 122GB RAM, GMKtec, EVO-X5 Pro

Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen…

Chinese motherboard maker borrows cable tech from enterprise servers to eliminate graphic card slot pave way for more external GPUs — Maxsun uses MCIO to fast-track compact workstations

1 outlet · 1 article · MCIO, data protection, cybersecurity, technology policy, privacy

Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master's and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen…

GUTcert critical infrastructure data breach

1 outlet · 1 article · ISO 27001, incident response, credential rotation, Kerberos, unauthorized access detection

A GUTcert critical infrastructure data breach has exposed roughly 640 GB of documentation submitted by German energy-grid operators as certification evidence, including exact substation locations and network topology maps. On September 24, Darmstadt utility ENTEGA AG confirmed…

Anthropic Could Start IPO Roadshow Week of Nov. 9, Putting Debut Before Thanksgiving

5 outlets · 5 articles · 12 total since September 7, 2026 · anthropic, bloomberg, Oura, SpaceX, OpenAI

Anthropic could begin formally marketing its initial public offering as soon as the week of Nov. 9, putting its shares on track to trade before Thanksgiving, Bloomberg reported Thursday . The company has issued invitations for an Oct. 14 investor day at its San Francisco…

More coverage: impli.me · itdaily.fr · businessinsider.es · tbpndigest.com

Namibia Opens National AI Institute Built Around Four Gaps Its Own Assessment Found

2 outlets · 3 articles · Emma Theofelus, Namibia, Artificial Intelligence, data governance, software development

Namibia’s National Commission on Research, Science and Technology opened a National Artificial Intelligence Institute on 23 September, structured around four centres that correspond directly to weaknesses identified in the country’s own 2025 AI Readiness Assessment. That…

More coverage: iafrica.com