|
|
Hi there, this is your daily ☕️ Cyberpresso.
|
|
|
In today's Cyberpresso:
|
|
🤖 AI agent hacked AI hub Hugging Face 🧾 EY leaks client tax records 🩺 Cancer diagnostics unit hit by breach 🖥️ ServiceNow AI flaw now under attack 🔐 Chinese hackers breach DigiCert Plus: 💡 4 strategies & tactics, 🎁 2 other news you might like, 🧰 6 tools, and 📚 5 papers.
|
|
The browser that reads the room before you ask.
Most browsers get you to the page. Norton Neo gets you to the answer. Magic Box understands your intent before you finish typing — no prompting, no switching apps, no copy-pasting. Built-in AI, instantly and for free. Privacy handled by Norton, by default.
|
🤖 AI agent hacked AI hub Hugging Face
LINK
|
- Hugging Face confirmed an autonomous AI agent breached its production infrastructure during the week of July 13, forcing its security team to run China's open-weight GLM 5.2 model on their own systems after commercial AI guardrails blocked their incident response.
- The attacker abused two code-execution paths in Hugging Face's dataset processing — a remote-code dataset loader and a template-injection in a dataset configuration — to run code on a processing worker, then escalated to node-level access and stole cloud and cluster credentials.
- Hugging Face said the attacker moved laterally into several internal clusters over a weekend using short-lived sandboxes, told customers to rotate access tokens and review recent activity, and reported no tampering with models, datasets, or spaces while its data-impact assessment continues.
|
🧾 EY leaks client tax records
LINK
|
- Ernst & Young (EY) is notifying tax clients that their personal and financial information was exposed after attackers breached a third-party IT service management platform used by the firm's tax practice to handle support tickets.
- Between March 28 and April 12, 2026, an unauthorized third party accessed the platform and downloaded documents containing client personal information and financial data used to prepare tax filings; EY detected the anomalous activity on April 23, 2026.
- EY says the unauthorized access has been stopped and systems secured, and is offering affected individuals 24 months of Experian IdentityWorks credit monitoring, with enrollment required by October 31, 2026; the notice does not name the threat actor, the provider, or how the platform was compromised.
|
🩺 Cancer diagnostics unit hit by breach
LINK
|
- Abbott Laboratories confirmed a breach of internal legacy Exact Sciences systems in its Cancer Diagnostics business after the ShinyHunters extortion gang listed the company on its leak site, first threatening to publish data after July 18, then extending to July 21.
- ShinyHunters told BleepingComputer it broke in through a vishing attack on several Abbott employees in mid-June, compromising a Microsoft Entra single sign-on account to reach internal systems, then stealing data from connected apps like ServiceNow, SharePoint, Databricks, and Coupa.
- The gang claims it took more than 30 million rows of customer personal data, over one million Social Security numbers, 22 million client notes with doctor-patient conversations, and 20 million medical orders; Abbott says the incident does not impact operations and BleepingComputer has not verified the claims.
|
🖥️ ServiceNow AI flaw now under attack
LINK
|
- Attackers have started exploiting a critical flaw (CVE-2026-6875) in the ServiceNow AI Platform, the enterprise workflow platform formerly known as the Now Platform, threat intelligence firm Defused reported over the weekend.
- The bug lets attackers with no login break out of the sandbox and run code remotely inside the ServiceNow platform in high-complexity attacks, hitting the same pre-auth entry point (/assessment_thanks.do) documented by Searchlight Cyber but reaching code execution by a different route than the published proof-of-concept.
- Defused observed the first exploitation attempts on Friday, days after ServiceNow patched hosted instances and released updates for self-hosted instances on July 13th, though ServiceNow still states it is "not currently aware of exploitation" and urges customers to upgrade to a patched release.
|
🔐 Chinese hackers breach DigiCert
LINK
|
- A China-linked hacking group known as GoldenEyeDog breached DigiCert in April 2026, stealing legitimate code-signing certificates that let them make malicious files look trustworthy and slip past Windows SmartScreen protections.
- Attackers infected a DigiCert support employee's device using malware sent through the support ticketing workflow, then intercepted certificate initialization codes meant for real customers and used them to sign their own malicious payloads.
- The signed files deliver a remote access trojan called Golden Gh0st RAT that steals browser logins from Chrome, Firefox, and QQ Browser, runs remote shells, creates a hidden admin account for automatic RDP login, and erases Windows Event Logs.
|
|
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
Other
news & articles you might like
-
North Korean Contagious Interview Campaign Hides OTTERCOOKIE Malware in SVG Images
LINK
-
Hidden prompts can secretly rewrite an AI’s memory, and researchers say that’s a serious problem
LINK
|
|
🛠️ Trending tools
|
AutoCVE: automatically scans source code to find security vulnerabilities, verifies them, and generates detailed reports using an intelligent agent system.
LINK
|
|
Akashic: a self-hosted intelligence workspace for organizing, analyzing, and connecting data, inspired by Palantir's data integration platform.
LINK
|
|
Tarit: a hypervisor for running lightweight virtual machines, designed to boot and execute workloads roughly twice as fast as Firecracker.
LINK
|
|
Signal in the Dark: a Three.js-based browser game where you must find another person hidden in darkness by choosing to illuminate them or yourself.
LINK
|
|
Peek-CLI: a command-line tool that lets Claude Code capture screenshots of front-end designs, enabling iterative visual feedback during development.
LINK
|
|
Fallout1-CE-PSP: a port of Fallout 1 (using the community engine fallout1-ce) that runs natively on the PlayStation Portable handheld console.
LINK
|
|
📚 Trending research papers
|
Software supply chain snapshots recreate exactly which package versions and known security flaws existed at each past release across npm, PyPI, and crates.io, letting teams audit dependency risk retroactively instead of only in real time.
LINK
|
|
Delivery status tracking for outbound business texts and emails now double checks itself, using a backup scan to catch messages stuck "in progress" when a provider's real time confirmation silently fails to arrive.
LINK
|
|
AI agent glitches found that many failures happen only when an AI's specific reply triggers a bad reaction from the surrounding automation code, making these bugs silent, hard to reproduce, and slow for developers to fix.
LINK
|
|
App wiring design automatically generates and verifies the connective logic linking software features, and testing shows guiding it with example scenarios, rather than plain instructions, more reliably produces the app behavior developers actually intended.
LINK
|
|
Compiler blind spots get systematically hunted down by an AI that studies which code lines existing tests never touch, then writes new test programs aimed squarely at those gaps, catching bugs older testing tools miss.
LINK
|
|
|
|
💬 How did you find today's edition?
We read every reply — just reply to this email and let us know how we can improve!
|
|
