AI Weekly Malaysia

Apple's support documentation page details how to turn off and restrict access to Apple Intelligence features on Mac, covering macOS versions including macOS 27 Golden Gate, macOS Tahoe 26, and macOS Sequoia 15. The page is part of the broader Mac User Guide and sits alongside sections on Siri AI, ChatGPT integration, Writing Tools, Image Playground, and Genmoji.

Tom's Hardware highlights a pre-built gaming PC priced at $2,299 featuring an AMD Ryzen 7 9800X3D CPU, an RTX 5080 Founders Edition GPU (described as DLSS 5-capable), 32GB RAM, and a 1TB SSD. The article is essentially a product deal listing with no independent benchmarking or technical analysis.

A joint advisory from Japan, U.S., Australia, and Germany reports that North Korean threat actors behind the 'Contagious Interview' campaign have compromised 30,000+ devices across 100+ countries since at least 2022, stealing $10.71M from 7,000+ crypto wallets. The attackers pose as recruiters on LinkedIn, lure developers and Web3 specialists into fake coding assessments, and deploy malware families like BeaverTail, InvisibleFerret, and OtterCookie via the infection chain.

Ireland's Data Protection Commission fined Google €403 million for GDPR violations in how three features—Web & App Activity, Location History, and Location Accuracy—handled user location data from May 2018 to February 2020. The DPC found breaches in lawful/fair processing, transparency, and excessive data retention, and ordered Google to bring its processing into compliance within 6 months. The fine is the DPC's fourth-largest but cannot be collected until an Irish court confirms it; Google can appeal within 28 days.

The article title claims Google confirmed Gemini models hacked three companies in May 2026, but the retrieved page content contains only cookie consent boilerplate with no substantive article text, quotes, or technical details.

Newly unsealed statements in the New York Times lawsuit against OpenAI and Microsoft include an OpenAI executive allegedly admitting chatbots are an 'existential threat' to journalism, a Microsoft director describing training as 'the largest theft of labor in human history,' and internal data showing NYT click-through rates dropping. The filings threaten the fair-use defense by suggesting AI products are direct substitutes for the source work.

Oura's upcoming $2.2B IPO is primarily a liquidity event for existing shareholders, with 36.5M of 50M shares offered coming from selling stockholders. Forerunner Ventures alone plans to sell its entire 9.3% stake for ~$1.2B at the $42 midpoint price, while Oura itself nets only ~$6.2M for general purposes after using $526.4M to cover tax obligations on vesting employee grants.

The Trump administration reportedly plans to give Russel Vought, a political appointee, veto power over all NIH grants. The article body was not accessible—only cookie consent boilerplate was captured—so details beyond the headline are unavailable.

Former accountant Ahad Ali built Tabby, an AI-powered real-time bookkeeping product that uses Plaid to import live account data and auto-generates financial dashboards, aiming to make the visible accounting software layer 'nonexistent.' After 14 months, Tabby has 5,500 small businesses and roughly $100K ARR with a team of just seven people, targeting end users directly rather than through accounting firms.

The article body was not captured — only cookie consent boilerplate from Ars Technica is present. No substantive content about the California settlement, the Paramount/Warner merger, or advocate reactions is available in the provided text.

Lon Lundgren measured a six-week decline in 'Fable 5' reasoning tokens after Anthropic made it permanently available in subscriptions, finding August delivered dramatically fewer thinking tokens than July across five measurement methods. Even at max effort levels, most invocations received little to no thinking tokens, and longer runs rarely matched published benchmarks.

Local opposition blocked 45 data center projects worth $68 billion in Q2 2026, even as overall data center investments are reportedly still on track to hit $32 trillion by 2050. The article provides no further substantive detail beyond the headline figures.

Volt Auto, the Malaysian distributor for Dongfeng passenger vehicles, confirmed that the Voyah Dream 9 MPV will launch in Malaysia in Q1 2027 with both BEV and PHEV variants. The BEV has a 120kWh battery with 701km CLTC range, the PHEV offers 300km WLTP pure-electric range, both on 800V architecture with 20-80% DC charging in 12 minutes. The vehicle includes Level 3 autonomous driving via Huawei's Qiankun ADS 5.0 system.

xAI announced Grok 4.7, a new larger base model trained with a longer RL run weighted toward multi-hour tasks. It is priced at $2/M input and $6/M output (same as Grok 4.6), scores 46.3% on CursorBench 4.0 and 71.0% (high effort) on DeepSWE v1.1, and ships with a new safeguard stack that tops LatchBio biosafety at 62.4% and allows only 3.3% of risky prompts on HackerBench v0.3.

AMD's stock price rose enough to push its market capitalization above $1 trillion, joining Nvidia, Broadcom, and SK hynix in that tier while Intel has not yet reached it. The article text itself is almost entirely boilerplate with no additional financial or technical detail beyond the headline.

The article title indicates Trump has rejected calls for an AI slowdown and instead launched an initiative called "AI Force." However, the article body was not captured—only cookie consent boilerplate is present, so substantive details about what "AI Force" entails are unavailable from this source.

Claire reviews Meta's Muse, a consumer AI agent that manages calendars, email, goals, and browser tasks, finding its UX notably polished—no terminal exposure, no awkward permission interruptions, and a first-attempt family newsletter that outperformed OpenClaw and Codex. Muse's permission pattern asks at the moment access becomes relevant, summarizes what it learned, and confirms before acting; its activity feed logs every tool call, script, and step for transparency. The Warp segment on shipping 2,000 PRs/month with AI factories is mentioned in the title but not detailed in the excerpt.

TechCrunch announces a Disrupt 2026 main stage session featuring all five Benchmark general partners—Jack Altman, Peter Fenton, Chetan Puttagunta, Everett Randle, and Eric Vishria—discussing where the next breakout startups will come from. Benchmark recently raised ~$2B across a $750M flagship fund and its first $1.25B growth fund, a notable shift for a firm known for concentrated early-stage investing. The article cites OECD data that AI companies captured 61% of global VC in 2025 ($258.7B of $427.1B), with deals over $100M making up ~73% of total AI investment value.

AMD's budget Ryzen 5 5500F and 7500 CPUs have appeared at retail priced roughly $20 above their announced MSRP, even from first-party sellers. The article text itself is mostly site navigation and membership boilerplate, so concrete retail context beyond the headline is thin.

Google's $899 Googlebook, now available for pre-order, runs Android OS with a desktop Chrome browser and bakes Gemini AI directly into the OS experience. Key features include 'Magic Cursor' (an AI-powered pointer that lets you highlight content for Gemini to act on), 'vibe-coded widgets,' and 'Rambler,' an AI dictation feature that cleans up rambling speech into readable text. Google positions it as a Chromebook successor that could migrate millions of users to Gemini-native computing.

Huawei has shelved the global rollout of its AI chips because domestic Chinese demand already exceeds supply. The title also references 15,488-chip Atlas clusters using optical networking to compete with Nvidia, scaling to 120 EFLOPS.

TechCrunch announces a fireside chat at Disrupt 2026 (Oct 13-15, San Francisco) with Google Search VP of Product Robby Stein, titled 'From MVP to Billions of Users: How Product Decisions Must Change at Scale.' The session will explore how product teams balance speed with trust and innovation with reliability as user base grows.

A personal essay arguing that algorithmic recommendation feeds on platforms like YouTube, Spotify, LinkedIn, and Reddit are hijacking attention and shaping thought, with AI slop and LLM-generated content now polluting these spaces. The author advocates returning to an 'intentional internet' where users actively choose destinations via bookmarks rather than passively consuming algorithmic suggestions.

A weekly security recap covering a CVSS 10.0 Cisco ISE auth bypass (CVE-2026-76460) under active exploitation, AI agent remote code execution vulnerabilities, surging ClickFix social-engineering attacks that trick developers into running malicious code, and a case where Hacktron used Anthropic's Claude Opus 5 to chain two critical vulnerabilities (including an SSO misconfiguration) against OpenAI.

Securonix researchers disclosed a campaign called TASK#STOMP that delivers a PowerShell backdoor via an encoded VBScript file on the victim's desktop. The backdoor exfiltrates business documents, Wi-Fi passwords, clipboard contents, and screenshots, using scheduled tasks disguised as legitimate OS services (e.g., 'Local Credential Manager', 'Network Audio Service') for persistence, plus a backup method via the Windows Startup folder. Initial access is unclear but likely phishing or social engineering.