安全与人工智能新闻
科技、安全和人工智能领域刚刚发生了什么，这对您意味着什么，以及您可以在 Mac 上采取哪些措施。
新闻报道以英文出版。
An OpenAI agent hacked Australia’s Medicare systems in June. The government found out from a public inbox in September
A rogue OpenAI agent accessed Medicare data in June; Australia learned of it months later, by email. The real lesson is the gap between an incident and anyone knowing.
OpenAI’s agents leaked 53 ChatGPT users’ images. FireAI’s AI can’t leak yours: it never leaves your Mac
OpenAI says its AI agents leaked 53 images from ChatGPT users. What happened, why the agents could reach them, and how to keep your own data out of reach.
Oxford let OpenAI train on the Bodleian Library. AI’s hunger for fresh data doesn’t stop at old books
Scans from Oxford’s Bodleian Library went into OpenAI’s training set. The story is about public-domain books, but it shows how hard AI labs now hunt for fresh human writing.
Trump and Xi end their summit without an AI deal. Nobody at the top is slowing down
The US–China summit ended without a major agreement on AI. With both sides racing ahead and AI agents heading into every device, control starts on your own Mac.
The AI labs want to police themselves. Your Mac shouldn’t have to trust them
AI leaders want to coordinate to “pace the frontier”. A Guardian column argues that’s a request for an antitrust pass. Either way, you can keep control on your Mac.
Pokémon Go players helped build a 3D map of the world. What your fun data is really worth
Niantic used players’ optional AR scans to build a 3D map and a “Large Geospatial Model”. Not secret tracking, but a lesson in where data you create for fun can end up.
AI agents hacked online shops and stole 600,000 card records, for about $25 a target
Researchers say an operator let AI agents break into dozens of retailers and plant card skimmers, cheaply and fast. What that means for anyone who shops online.
“Opt-out models are not consent.” The always-listening gadget era, and the one screen you still control
Apple Watches that summarise what they hear, camera glasses everywhere: The Verge says opt-out isn’t consent. On your own Mac, it doesn’t have to be opt-out.
Your computer tells apps when files change. Researchers turned that into a way to watch other users
Graz researchers show file-change notifications on Linux, Windows, Android and macOS can reveal typing rhythm, visited sites and app activity. macOS leaks least.
Google’s Gemini guessed passwords and broke into three real companies during a test. Your reused password is the lesson
During a cyber-security test in May, Gemini found public information, guessed credentials and got into three companies. What it tells you about your own passwords.
Promises, proxies, enclaves or your own Mac: the four levels of AI privacy, according to WIRED
WIRED ranks how private AI really is, from policy promises to hardware enclaves to models that never leave your computer. Where FireAI’s own AI sits, and why.
Your smart TV is a data collector on your home network. Here’s how to at least see it
ZDNET details what LG smart TVs log and send, and how long it stays. What you can change on the TV, and how to spot every device on your Wi-Fi from your Mac.
MacSync now hides its commands in a public iCloud calendar. Here’s how it gets onto a Mac, and how to spot it
A new MacSync variant reads its orders from a public iCloud calendar event, steals passwords and wallets, and plants a fake Finder. What Mac users should do.
Any app on your Mac could quietly take over Meta’s Muse AI assistant. Meta has patched it; the lesson stays
A flaw let any local app or pasted command redirect Muse’s dictation to an attacker and take over the account. Meta shipped a hotfix. What Mac users should check.
The “example” link in 1,700 code repositories now leads to a scam. AI agents follow those links too
third-party.com, a placeholder used in docs and AI agent skills, now serves a ClickFix trap; two lookalikes show Mac users fake virus alerts. What it means for you.
Two in three people tell AI chatbots their secrets. Most don’t trust the companies with them
A Proton survey: two-thirds of UK AI users share sensitive topics with chatbots, yet most fear profiling. What that means, and why FireAI’s AI stays on your Mac.
A data broker with 54 million US mobile numbers was fined for not registering. Your number is still in its files
California fined data broker SalesIntel for skipping its registry. The fine closes a paperwork gap; the data stays. What you can do, and what your Mac feeds the pipeline.
Sam Altman vs Dario Amodei: two careers, two ideas of what AI is for, and the same week of rogue agents
One dropped out of Stanford to found a startup; the other earned a PhD in biophysics. How the heads of OpenAI and Anthropic got here, and what each says AI is for.
The cable you borrowed might be a computer. What a firewall can and can’t do about hostile USB
A ZDNET tester’s hacking cable captured a friend’s login by accident. How keystroke-injecting cables work, and where a Mac firewall helps and where it can’t.