#безопасность
-
Vulnerability in Amazon's Kiro AI Editor Allows Data Theft via Prompt Injection
Researchers at Mindguard have shown how a malicious repository can force the agentic development environment Kiro to quietly send local data to an external server — without a single command from the user.
-
OpenAI Shut Down a Cambodian Fraud Network That Was Using ChatGPT
The group ran several types of scam at once, using ChatGPT to generate fake identities and forged documents. Bruce Schneier reported on it.
-
Two TeamPCP Suspects Arrested in Australia
The Australian Federal Police has arrested two young men from Western Australia believed to be members of TeamPCP — the group behind the longest-running string of software supply-chain attacks. The collective's leader was undone by his own carelessness.
-
The Carhartt Data Breach: 12.9 Million Accounts Compromised
The extortion group ShinyHunters has published the data of nearly 13 million customers and employees of the American clothing maker Carhartt after the company refused to pay a ransom.
-
CISA Adds Six Exploited Vulnerabilities to Its KEV Catalog, Including a Citrix NetScaler Flaw
The U.S. agency CISA has added six new entries to its list of actively exploited vulnerabilities — ranging from a fresh Citrix NetScaler flaw to old holes in Linux, SQL Server and Red Hat utilities.
-
ATF Confirms “Major Incident” Following Qilin Ransomware Claims
The U.S. Bureau of Alcohol, Tobacco, Firearms and Explosives has acknowledged a breach of one of its systems after the Qilin gang added the agency to its leak site.
-
Critical Avada Theme Vulnerability for WordPress: Zero-Click RCE
A chain of six vulnerabilities in the popular Avada theme lets an unauthenticated attacker execute arbitrary PHP code on the server. The flaw was found by a Wordfence AI tool in under two hours.
-
OpenAI Publishes Report on the Hugging Face Breach Carried Out by Its Own Model
The company has presented the most complete account to date of the incident: during testing, an AI model broke out of its test environment and, step by step, compromised systems belonging to OpenAI, Hugging Face and third-party vendors.
-
GPUThor: New Rowhammer Attack Defeats ECC Protection on NVIDIA GPUs
Researchers at the University of Toronto have demonstrated an attack that bypasses memory error correction on NVIDIA's Ampere-family GPUs, enabling both denial of service and privilege escalation to root.
-
SLEEPWALKER: the backdoor that sleeps until one “magic packet” arrives
An independent researcher has described a previously unknown Windows backdoor that lies dormant in memory until it receives a specially crafted network packet, then executes commands written in a custom 23-instruction language.
-
AI agent “Alice from Apple Support” tricks theft victims into handing over their device passwords
Researchers at SOCRadar have detailed the AnonyMousKIT platform — a “phishing as a service” operation that uses rented AI voice agents to call the owners of stolen iPhones in Apple's name and coax out their passcode, Apple ID credentials and one-time 2FA code.
-
LACMA Data Breach: Social Security Numbers and Medical Information at Risk
The Los Angeles County Museum of Art has disclosed that a 2025 breach of its networks gave attackers access to the personal data of visitors and staff — from passport numbers to medical diagnoses.
-
Attack on Zimbra: Over 270 Servers Breached Through an SNMP Vulnerability
Attackers are actively exploiting the critical vulnerability CVE-2026-73570 in Zimbra Collaboration Suite, which allows remote code execution without authentication. According to Shadowserver, hundreds of servers have been compromised.
-
Attacks on the miniOrange SAML SSO plugin for WordPress: login bypass grants admin rights
Attackers are mass-scanning WordPress sites for the Xecurify miniOrange SAML 2.0 SSO plugin, whose two vulnerabilities allow logging in as any user without authentication.
-
CISA Adds Critical Oracle WebLogic Vulnerability to Its Catalog of Actively Exploited Flaws
The CVE-2026-21962 vulnerability, with a maximum CVSS score of 10.0, lets an unauthenticated attacker access critical data over HTTP. Patches shipped back in January, but the attacks continue.
-
Can an AI Take Over the Server Through Inference-Engine Vulnerabilities?
An essay by Boyd Kane on how a malicious language model could gain control of the very machine hosting its own weights by exploiting bugs in engines like vLLM and SGLang.
-
Vulnerability in Calix routers exposes home devices to the entire internet
An unpatched hole in Calix GS7 XGS routers lets attackers create port-forwarding rules remotely and without authentication, bypassing NAT and the firewall. The manufacturer never responded to the researcher's outreach.
-
Alabama Investigates Incident Involving OpenAI Model That Hacked Hugging Face
The state's attorney general has served OpenAI with a subpoena, seeking to determine whether the company violated consumer protection laws. The trigger was a cyber model that broke out of an isolated environment and attacked the Hugging Face platform.
-
Critical Keycloak Vulnerability Lets Attackers Hijack Any Account via Password Reset
Red Hat and the Keycloak project have released patches for the critical flaw CVE-2026-18963 (CVSS 9.1): an unauthenticated attacker can take over any account, including an administrative one, by forcing the system to reset the password.
-
UAT-10147: cybercriminals turn server attacks into an assembly line with AI
Cisco Talos researchers have profiled a Chinese-speaking group that mass-compromises Windows and Linux servers, leaning on AI tools and a new cross-platform SPECTRE implant with EDR bypass and a Linux rootkit.
- More