News
-
PoeLLM Hides Its C2 Address in a Poem, Infecting 3,400+ Servers
The Canto Incognito campaign hunts internet-exposed AI and LLM services, turning them into cryptocurrency miners and scanners that seek out fresh victims.
-
Unpatched LMCache Flaw Hands Remote Code Execution to Anyone on the Network
JFrog has disclosed a critical hole in LMCache, the open-source accelerator for LLM servers: a single network message can run code on the cache server — as root on the project's official images. No fix exists yet.
-
Mistral unveils "Le Chonk," an open-weight challenger to the top models
French AI company Mistral says its new model, Le Chonk, can rival the best closed systems while staying open-weight.
-
ShinyHunters Leader Nabbed Mid-Extortion of Boeing Spin-off
A teenager from Amman who led the data-theft gang has been detained and, per Reuters, is now cooperating with the FBI. He was caught while his group was extorting a former Boeing unit, Jeppesen ForeFlight.
-
FBI: FortiBleed Harvested 86,644 Fortinet Device Credentials and Is Still Active
The FBI and U.S. Secret Service warned that the FortiBleed credential-harvesting campaign remains an active threat to internet-facing Fortinet FortiGate firewalls and SSL VPN gateways.
-
Apple Unveils Reference Image, a Verified Photography System
Apple has released Reference Image, a system that confirms a photo is exactly as an iPhone captured it — without revealing the specific device or the photographer behind it.
-
Anthropic Opens Claude to Vetted Cyber Teams as Glasswing Flags 129,000 Flaws
The company is widening access to its models with relaxed guardrails and reports tens of thousands of AI-found software defects. Yet very few of them prove exploitable in practice.
-
SonicWall patches max-severity SSRF flaw in SMA1000 gateways
The company has issued emergency hotfixes for its SMA1000 series: the flaw let an unauthenticated attacker make the appliance issue requests on their behalf and reach internal functionality.
-
CERT-UA: 100+ Hacked Sites Push LunexStealer via Fake Cloudflare Check
Ukraine's CERT-UA has found over a hundred compromised websites that, under the guise of a "prove you're human" check, coax visitors into running a malicious command themselves.
-
When random isn't random enough: the modulo trap
The classic way to pick a random object — grab a big number and take it modulo the count — quietly breaks uniformity. The author explains why, and what to reach for instead.
-
Raspberry Pi Desktop returns for PC and Mac — now on Debian Trixie
After a long pause, Raspberry Pi has refreshed its desktop environment for ordinary computers: the new build runs on 64-bit Debian Trixie and ships with the Raspberry Pi Connect client.
-
Another Dozen Vulnerabilities Turn Up in X.Org Server and XWayland
With help from artificial intelligence, a fresh batch of twelve security flaws has been disclosed, affecting both the classic X.Org Server and XWayland.
-
Psiphon Ready to Leave Canada Over Lawful-Access Bill
Psiphon, a maker of censorship-circumvention tools, says it will leave Canada if Bill C-22 passes, forcing providers to secretly build surveillance capabilities for police and intelligence services.
-
OpenAI drops another batch of mathematical breakthroughs
The company has published 722 manuscripts containing solutions to long-standing open problems, produced by an unreleased frontier model — and once again rattled the mathematical community.
-
Hackers obtain counterfeit TLS certificates for Google and other services
Attackers managed to issue unauthorized TLS certificates for Google and several other large services after compromising three domain registries.
-
Why Embedding Models Should Be Open
Simon Willison explains why an open license matters more for embedding models than for many other AI systems: the Apache 2.0 license on EmbeddingGemma 2 is insurance against the model simply disappearing.
-
Linux Backdoors Pose as Email Security Tools to Hide in Korean and Taiwanese Networks
Rapid7 has detailed a family of Linux implants that disguise their traffic as email services and name their processes after SpamSniper and ShareTech, two email-security products popular across the region.
-
Pwn2Own Ireland 2026: 32 zero-days and $388,500 on day one
On the opening day researchers cracked the Samsung Galaxy S26 flagship twice and earned nearly $390,000 by exploiting 32 previously unknown flaws.
-
Atlassian patches critical file-access flaw in Jira and Confluence
The CVE-2026-21589 vulnerability lets an unauthenticated attacker read specific files in a web application's root directory. It affects self-hosted Data Center products, including Confluence, Jira and Bitbucket.
-
OpenAI agents tried to hack Wikipedia's tools and flooded it with traffic
According to Ars Technica, OpenAI's autonomous agents have harmed yet another third-party site — this time Wikipedia took the hit.
- More