1Highlight: Binance's RM25.6 billion theft began with a fake AI coding tool and a fake job test, and a tampered Claude Code add-on hit Cursor's store.
2How companies use AI: Heimstaden cut listing photo costs by 99%, Norton's team shipped 3x faster, and Anthropic's Rust rewrite cut computer use 40-50%.
3Also: four checks to stop AI workslop, an EU court adviser on Gemini and news, Itch.io's No AI filter, and Estek's RM20 million Penang plant.
Highlight of the day
AI and security
RM25.6 billion
stolen from Binance on 29 Oct (about US$6.3 billion)
Merged from 6 sources
★ Highlight of the day
Hackers used fake job tests and AI tools to rob Binance
Binance says the record theft of about RM25.6 billion (US$6.3 billion) on 29 October began months earlier with a fake AI coding tool and a fake job interview. In the same week, a tampered Claude Code Extension A small add-on you install into an app, such as a browser or a code editor, to give it extra features. sat in Cursor's app store for about 11 hours.
What we learned
How it started: in June, a Binance engineer looked for a popular AI coding tool and installed a tampered copy from a fake download page. In September, in a Social engineering Tricking people, rather than computers, into giving access or running something harmful, for example with a fake job offer. trick, fake recruiters gave another engineer a coding test with hidden Malware Harmful software that steals data, spies on you or damages your computer.. The attackers then moved through Binance's own internal AI tools to the system that approves money transfers (Binance's report, via CNBC).
Binance says it has put back 98% of the money from its own funds, and about RM11.6 billion (US$2.85 billion) has been frozen or seized. The FBI blames a North Korean hacking group called TraderTraitor. North Korea denies it.
On 29 October, a tampered copy of the Claude Code extension was live in Cursor's extension store for about 11 hours and reached at most 5,000 users. It tried to plant a hidden Backdoor A hidden way into a computer or system that lets an attacker get back in later without the owner knowing. into their computers, a classic Supply chain attack A hack that hides harmful code inside a tool or update people trust, so it reaches many users at once.. Removing it is not enough: Cursor says to update to version 2.1.314 or later (Cursor, WIRED).
AI also finds Vulnerability A weakness in software that an attacker could use to break in or cause harm. faster. Anthropic says an unreleased Claude model found and used a Linux bug that gives an attacker Root access Full control of a computer, able to read, change or delete anything on it., or full control, on some servers, in about 53 minutes by Fireship's count. A fix is out.
Close to home: Prime Minister Anwar warned on 3 November that Malaysia could be a target for cyberattacks, and Bank Negara is meeting the banks about it (Lowyat.net, citing The Edge).
Why it matters: Attackers now go after the tools people install to use AI. Download AI tools only from the official site, be wary of coding tests from recruiters you don't know, and keep updates switched on.
Binance co-CEO Teng says hackers used fake job interviews to steal $6.3 billion in crypto · CNBCRead ↗
North Korea responsible for $6 billion Binance attack · FBIRead ↗
Claude Code extension security notice · CursorRead ↗
Trojan malware targets thousands of Anthropic Claude developer systems · WIREDRead ↗
Fireship on AI-found Linux bugs and recent hacks · FireshipWatch ↗
Anwar: Malaysia could be a target for cyberattacks · Lowyat.netRead ↗
Was this useful?Tap to vote. Hold, right-click or press Shift+Enter to add a reason or a note first.
More stories today
How companies use AI
99%
lower cost per listing photo at Heimstaden in the Netherlands
Merged from 3 sources
How companies use AI
A landlord, a security firm and Anthropic measure AI's payoff
A European housing landlord cut the cost of listing photos by 99%. Norton's maker built a news app three times faster, and Anthropic rewrote a core system in a language its team didn't know.
Key learnings
Heimstaden (rental housing, Europe): a listing photo with furniture added by computer (Virtual staging Adding furniture and decor to photos of empty rooms by computer, so property listings look lived-in.) used to cost about RM228 (€50). With AI it now costs under RM0.46 (€0.10), which the company estimates could save about RM590,000 (€130,000) a year in the Netherlands alone (Heimstaden, in OpenAI's case study).
At Heimstaden, an analyst with no coding background used Codex OpenAI's AI coding helper, which can write, fix and test software for you. to check 20,112 properties for climate risk. The company says this took 2 people two months, against an estimated 5 people for a year. Half its staff used ChatGPT within 2 to 3 months of launch.
Gen Digital (maker of Norton): a small team built the Norton Revamp news app, launched in March 2026, with AI coding help. It shipped features 3 times faster and about 70% of the screen code was AI-written. But checking the AI's work took 20 to 30% of their time (Gen's engineers, AWS blog).
Anthropic: a team with no Rust A programming language known for fast, safe code, often used where speed and reliability matter. experience used Claude to move a core part of its API Application programming interface: a door that lets one piece of software ask another for data or actions. from Python to Rust. Computer use When an AI controls a computer like a person does: looking at the screen, clicking and typing to get a task done. fell 40 to 50% and first replies came 10 to 20% faster, says Anthropic's Katelyn Lesse (InfoQ).
Lesson for small firms: the biggest savings come from repeated jobs, like photos or property checks. Budget real time for someone to check what the AI produces.
Why it matters: These are measured results, and Norton's team also counted the cost of checking AI's work. That hidden cost is what a small business needs to plan for.
Heimstaden scales AI across its business with ChatGPT and Codex · OpenAIRead ↗
How Gen Digital built Norton Revamp's AI-curated news engine with Amazon Bedrock AgentCore · AWS Machine Learning BlogRead ↗
Anthropic moves its Messages API from Python to Rust with Claude (Chinese) · InfoQ ChinaRead ↗
Was this useful?Tap to vote. Hold, right-click or press Shift+Enter to add a reason or a note first.
Using AI well
~2 hours
lost per piece of sloppy AI work received (Stanford/BetterUp)
Merged from 2 videos
Using AI well
How to stop sloppy AI work from wasting your colleagues' time
Sloppy AI-made work, called 'Workslop AI-made work that looks finished but is vague or wrong, so the person who receives it has to fix it.', costs the people who receive it hours to fix. The fix is simple: give the AI your own view and some Context window How much text an AI can keep in mind at once, like the size of its short-term memory. before it writes anything.
Key learnings
A Stanford and BetterUp study found about 40% of US office workers had received workslop in the past month, and each case took close to two hours to deal with (The AI Daily Brief).
Four checks before you send AI-made work: could you defend every sentence? Did you read all of it? Does it include something only you could add? Would you put your name on it?
Say what you think first. The same plain prompt gave almost identical LinkedIn posts in ChatGPT, Claude and Gemini. Adding the writer's own opinion, and words to avoid, gave a much better draft.
Give it context once: Notion shows a setup that writes an 'operating guide' page. You paste it into ChatGPT's or Claude's Custom instructions Standing notes you give an AI chatbot about you and how to answer, which it then follows in every chat., so the AI knows where your projects and notes are, and you can save repeated jobs as Skill A saved, reusable set of instructions (and sometimes code) that an AI agent loads when a task needs it. (Notion).
Why it matters: AI makes drafts cheap, but the reader pays for the checking. A few minutes adding your view saves everyone else time and protects your name.
How to Deslopify Your AI · The AI Daily BriefWatch ↗
Set up ChatGPT and Claude with your Notion workspace · NotionWatch ↗
Was this useful?Tap to vote. Hold, right-click or press Shift+Enter to add a reason or a note first.
AI and creators
6,000+
new game pages a day on Itch.io, double the August level
Merged from 4 sources
AI and creators
EU court adviser: AI answers copying news articles can break copyright
An adviser to the EU's top court says Google's Gemini may break publishers' rights when its answers stand in for the original article. Separately, the game store Itch.io deleted terabytes of AI-made junk.
Key learnings
In the EU court's first case on generative AI, Hungarian publisher Like Company v Google, Advocate General A senior legal adviser at the EU's top court who gives judges a non-binding opinion before they decide a case. Szpunar said on 3 November that Gemini may break the rules if it gets past Paywall A barrier on a website that lets only paying subscribers read the full content. or bot blocks, or if its answer copies an article so closely that readers no longer need the original (CJEU press release).
Facts are free to use; only the way a story is written is protected. He also said training AI on articles is generally allowed under the EU's Text and data mining exception An EU copyright rule that lets computers analyse large amounts of published text and data, for example to train AI, unless owners opt out.. His opinion is not binding, and the judges rule later (Courthouse News, RTÉ).
Itch.io: new game pages a day doubled from about 3,000 to more than 6,000 between 3 August and 30 October, but downloads did not rise. The store deleted 'a couple terabytes' of AI slop Low-quality content made quickly with AI and passed on without proper checking. and added a 'No AI' search filter (Kotaku).
The filter only shows games whose makers say they used no AI, so most games now drop out of it. Makers of tabletop games, books, comics and soundtracks must now say whether they used AI.
Why it matters: Courts and stores are starting to set limits on AI reusing people's work. If you use AI to write or make things, credit and link the original, and say when AI helped.
Advocate General's Opinion in Case C-250/25 Like Company v Google (press release) · Court of Justice of the EURead ↗
EU legal adviser finds limited uses of Google AI could run afoul of publishers' rights · Courthouse NewsRead ↗
EU court adviser backs publishers in Google AI case · RTÉRead ↗
Itch.io deletes terabytes of gen-AI slop and adds a No AI filter · KotakuRead ↗
Was this useful?Tap to vote. Hold, right-click or press Shift+Enter to add a reason or a note first.
Southeast Asia
RM20 million
Estek's new AI and chip-equipment plant in Batu Kawan, Penang
Merged from 6 sources
Southeast Asia
Penang firm opens RM20 million plant for AI chip inspection machines
A homegrown Penang company opened a new plant that builds chip-making machines and AI inspection tools. Singapore's NCS is tripling its AI deployment engineers, and a Thai bank's fund backed AI group Amity.
Key learnings
Malaysia: Estek Automation, a Penang company started in 2015, opened a 50,000 sq ft plant in Batu Kawan worth RM20 million. It makes chip-making machines such as die sorters and Wafer inspection Checking the thin discs of silicon that computer chips are made on, often with cameras and AI, to catch faults early., and AI camera checks that spot faulty parts (The Star, Digital News Asia).
Singapore: NCS opened an AI Mission Centre to help organisations move from AI trials to daily use. It plans to grow its embedded AI engineers from 100 to more than 300 in two years, and signed deals with Changi Airport Group, the Institute of Mental Health and Life3 Biotech (Digital News Asia, CNA).
Thailand: Krungsri Finnovate, the venture arm of Krungsri bank, invested in Amity, extending its Series D round past about RM407 million (US$100 million). Amity serves more than 100 business partners and thousands of small businesses, and is eyeing a stock market listing (TechNode Global, e27).
Why it matters: In the region, AI money is going into factories and the people who put AI to work, not only chatbots. That means jobs for engineers and technicians, not just coders.
AI and security
