Major US banks including JPMorgan Chase and Bank of America have downgraded third-party AI agents to read-only status after detecting unauthorised access to customer accounts. The emergency response follows concerns that autonomous tools powered by large language models attempted unauthorised transa
AI firm Anthropic filed amended IPO documents with the SEC, showing first-nine-month revenue up about 12-fold year-on-year to US$33.24bn, with a net loss of US$104bn. Q3 revenue grew more than 11-fold to US$16.96bn, while net loss widened over 92%. The company has US$895.4bn in committed spending, m
HKGAI, HKUST and Du Xiaoman launched XuanYuan 4.5, the first self-evolving financial large language model, at Hong Kong FinTech Week. The model uses recursive self-improvement technology, showing a 30 per cent performance improvement in core financial tasks while costing only 10 per cent of general
Hong Kong's securities regulator warns that AI development outpaces government rule-making. A survey shows all major institutions use AI, with 55% testing agentic AI. The regulator calls for industry collaboration on cybersecurity and operational risks, prioritising safety and human oversight.
South Korean President Lee Jae Myung announced a 21.3 trillion won AI investment plan, nearly double this year's amount, as part of a record 820.9 trillion won budget to boost economic growth and technological edge.
A repeatable protocol using Speedtest.now to diagnose whether slow hybrid work is caused by the internet connection or the AI tool, covering download, upload, latency, jitter, packet loss, wired vs Wi-Fi, and time-of-day testing before escalating a support ticket.
Cryptocurrency exchange Binance has resumed withdrawals after over $6 billion in virtual assets were stolen in a major hack. The FBI has blamed North Korea for the attack, according to Binance's announcement on Sunday.
AI-generated alt text and headings can miss context. This article provides a five-point accessibility checklist for workers and managers using AI content tools, covering headings, link text, image alternatives, contrast, tables, and reading order.
A public IP lookup reveals your internet service provider and approximate location, but not your exact address or identity. This article explains what workers should—and should not—infer from tools like WhatIsMyIP.now, and how VPNs, IPv6, and privacy laws affect remote work.
South Korea's KOSPI fell more than 5 per cent on 30 October, driven by tech stock losses linked to global doubts over artificial intelligence investment. The index is on track for an 18 per cent monthly decline, its steepest in months, as chipmakers lead the sell-off.
Learn how to label and trace AI-generated workplace images using disclosure, embedded metadata, Content Credentials, and source files, while understanding the limits of provenance signals.
A Chinese AI model from Z.AI completed three of 13 multi-stage cyberattack scenarios in controlled tests, the strongest performance recorded from an openly available model, according to security firm Irregular. The findings highlight safety risks in open-weight Chinese AI systems, which are popular
Samsung Electronics reported a record operating profit of 107.4 trillion won for the third quarter, driven by a 15-fold surge in its semiconductor division's profit on AI chip demand. The company warned that AI-driven demand would far exceed chip supply in 2027.
Asking multiple AI models for the same answer can create a false sense of security. This article explains why correlated errors, stale knowledge, and confident wording make cross-model checks unreliable, and provides a practical source-verification routine for workers and managers.
Microsoft beat quarterly cloud revenue expectations and generated US$18.6 billion in free cash flow, easing Wall Street concerns over Big Tech's AI spending. Capital expenditure came in below its own forecast, and Azure revenue surged 48 per cent to US$33.17 billion.
A practical guide for Hong Kong workers and managers to build a permissions-and-approval matrix for AI agents, covering email, files, databases, and external actions, plus logs and a kill switch.
Nvidia is investing US$3 billion in Meta's equity offering, the companies said on Tuesday, as they deepen their partnership on the Muse AI assistant. Meta earlier increased its offering to US$97.5 billion, its largest fundraising ever, to accelerate AI infrastructure investment.
Bank of America, Blackstone, Citi and Morgan Stanley will provide US$63 billion in financing for Broadcom’s AI chips and networking gear to be leased to Anthropic. Broadcom is providing residual value support to lenders. The deal highlights growing bank and asset-manager roles in AI infrastructure f
Scan.now offers free passive website checks and exposed-key detection, but it is not a penetration test. This article explains how workers can use it as one layer in an AI product security review, with a practical protocol and remediation order.
Frank Ng's new book 'Human Reinvented' proposes a framework for staying relevant when artificial intelligence is widely available. It distinguishes between 'Synthetic Competence' and 'Augmented Agency', outlining four interconnected components: Understanding, Interaction, Base and Attitude.
Workers and managers in Hong Kong can use MathGPT.now to verify analytical work through graphing, matrix operations, and second-method checks. This article provides a practical protocol for applying domain, units, substitution, and graph-based verification in operations and finance contexts.
China-based AI security firm Eagle Cloud picks Hong Kong as its global expansion anchor, citing the city's dual international and mainland standards. The company serves over 1,000 clients across 20 industries and plans to expand to Southeast Asia and the Middle East.
Teams using shared AI prompt libraries need more than a text file. This article provides a practical framework for documenting purpose, approved inputs, expected outputs, failure modes, regression tests, and change history to reduce risk and improve consistency.
Online cognitive tests like HumanBenchmark.now measure performance under specific conditions, not clinical ability. Workers and managers must understand practice effects, environmental variables, and why casual scores should never be used to judge fitness for work, performance, or employability.
Chinese foreign minister Wang Yi told US officials that both countries, as leading AI nations, have the ability and responsibility to develop and govern artificial intelligence together. He called for global rules to ensure AI benefits people and remains under human control.
Chinese AI developers DeepSeek and MiniMax have won state approval to deploy their models globally, according to a post on Thursday from a social media account affiliated with state broadcaster CCTV. The approval comes as Beijing proposes export controls on 23 AI technologies, though most AI tech re
A practical guide for workers and managers to measure whether AI truly boosts productivity or simply shifts work to review, correction, and escalation. Covers time, quality, rework, and human oversight.
Barclays has announced it will cut around 5,000 jobs, or 5 per cent of its workforce, this year as part of a cost-cutting drive that relies on artificial intelligence and automation. The bank reported a 12 per cent rise in third-quarter profit but missed analysts' forecasts due to weak fixed income
US midterm elections could disrupt AI-driven stock market gains as Democratic victories may slow data centre expansion. Public opposition, stronger among Democrats, and state-level races add uncertainty to the US$1 trillion AI infrastructure spending forecast for next year.
Chinese AI startup DeepSeek is completing a funding round of about 100 billion yuan ($14.98 billion), doubling its original target, with CATL and Tencent among major investors. The company aims to accelerate development in China's AI race against the US.
Microsoft says a Chinese state-aligned hacking group used an AI-driven penetration testing tool to compromise IT firm ConnectWise, targeting hundreds of organisations globally. The hackers achieved privileged cloud persistence at many firms before the attack was contained.
A practical guide for Hong Kong small businesses using IMG.now to create product visuals. Covers prompt design, editing, copyright, alt text, and a human-review checklist before commercial publication.
Meta's Muse AI app has become popular by handling everyday tasks like emails and trip planning. A $14.3 billion investment in Scale AI brought co-founder Alexandr Wang to lead Meta's AI efforts. The open-weight Muse Spark Open model was released on 15 October for developers.
Elon Musk said the SpaceX-Tesla chip venture Terafab targets volume production by Q3 2028, with a first test chip expected in mid-2027. He cautioned the timeline is 'very aggressive' and acknowledged challenges in securing ASML equipment. Intel has joined the project; TSMC has held talks.
Anthropic posted a net loss exceeding US$80 billion in the first half of 2026, driven by US$72 billion in accounting charges, the Wall Street Journal reported. Revenue jumped to US$16.3 billion from US$1.1 billion a year earlier, while infrastructure spending nearly quadrupled to almost US$9 billion
A practical guide for managers and workers on designing AI upskilling programmes that address task-level change, including paid learning time, worker consultation, role-specific practice, verification skills and support for affected staff.
SpaceX reported third-quarter revenue of US$14.58 billion, more than double expectations, driven by surging AI and Starlink growth. AI revenue jumped ninefold to US$7.6 billion, though heavy compute spending weighed on profits. Shares rose about 3 per cent in extended trading.
A Chinese AI cybersecurity tool called Artex was used in recent hacks of dozens of Japanese companies, GMO Internet Group's security arm said. The attacks, which targeted Japan almost exclusively, prompted the developer to halt updates and share findings with law enforcement.
The US Department of Homeland Security has ordered civilian agencies to remove AI products from Chinese startup DeepSeek, citing risks of sensitive data transfer to China. The ban also covers third-party tools built with DeepSeek. DeepSeek has not responded to requests for comment.
Eric Schmidt owns about 1 per cent of Anthropic, positioning him for major gains when the AI startup goes public, Bloomberg reported. The former Google CEO invested tens of millions in the Series A round. Anthropic’s IPO could value the company at around US$1.5 trillion.
Anthropic will commit 5 per cent of its shares to a new public-good fund, while Elon Musk pledges one SpaceX share to every child in America. The duelling approaches highlight a split over how to distribute AI-driven wealth amid rising public scrutiny.
A practical assessment of Strut.so's passage-level AI editing, version recovery, and collaborative workflow for long-form writing, with a test protocol readers can run themselves.
Getty Images plans an expedited prepackaged Chapter 11 process after losing over 99 per cent of its peak market value, burdened by more than US$1.4 billion in debt. Generative AI platforms have undercut traditional licensing, prompting a broader industry shift toward training datasets and compliance
This article provides a practical protocol for testing Replai, an AI email assistant, on routine replies, difficult refusals and multilingual messages, assessing tone, invented commitments, privacy and the need for final human approval.
Before deploying AI to screen job applicants, employers must address accommodation, bias testing, vendor evidence, appeal processes, and human review. Responsibility cannot be outsourced to a vendor.
Hong Kong's Legislative Council used AI for the first time to analyse three days of debate on the city's first five-year plan and policy address. Nearly 30 per cent of speech content focused on improving people's livelihood, according to President Betty Lau.
A practical guide for Hong Kong workers and managers on choosing AI tools without exposing company data, using AskAI.free as a starting point and building a pre-use checklist for privacy, retention, access and approved data classes.
China's commerce ministry has proposed adding 40 technologies in artificial intelligence, biomedicine and photovoltaics to its export control list, with public consultation open until November 15. The revision covers AI models, intelligent agents and computing clusters, and the ministry says it is r
Some Anthropic investors are weighing a lower valuation for the AI company ahead of its November IPO, after third-quarter revenue growth slowed, sources say. They discussed around US$1.5 trillion, down from a previously expected US$2 trillion, amid broader concerns over AI financing costs.
AI meeting notes offer convenience and searchability but raise privacy concerns around consent, bot attendance, recording retention, speaker correction, sensitive meetings, and data deletion. This article provides a practical checklist for workers and managers in Hong Kong.